Railway freight train adaptive asynchronous control method based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a railway freight train adaptive asynchronous control method based on artificial intelligence. The method comprises the following steps that train running data are collected through a train sensor, train running and track linkage processing and train running real-time positioning detection are carried out, and train running-track linkage data and train running real-time positioning data are generated; acquiring train driving planning data; carrying out train running schedule design based on the train running planning data and the train running real-time positioning data, and generating train running schedule data; and based on the train running real-time positioning data, the train running-track linkage data and the train running schedule data, train carriage running state and train traction linkage analysis and train self-adaptive asynchronous control instruction setting are carried out, and train self-adaptive asynchronous control operation is executed. According to the invention, self-adaptive asynchronous control of the train is realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based adaptive asynchronous control method for railway freight trains. Background Art

[0002] With the development of intelligent railways, freight trains are placing higher demands on traction distribution, car state linkage, driving positioning accuracy, and scheduling adaptability during operation. Existing control methods for freight trains often rely on fixed rule-driven control logic or centralized scheduling based on preset paths. Faced with complex operating conditions such as varying track curve radii, undulating slopes, and tunnel communication blind spots, freight trains struggle to adaptively and asynchronously adjust their operating states and control strategies. This leads to problems such as uneven car state distribution, delayed traction response, and accumulated positioning errors. These issues further contribute to unbalanced traction coordination among individual cars, excessive stress fluctuations within cars, and uncontrolled driving rhythm. Summary of the Invention

[0003] Based on this, the present invention provides an artificial intelligence-based adaptive asynchronous control method for railway freight trains to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, an artificial intelligence-based adaptive asynchronous control method for railway freight trains comprises the following steps:

[0005] Step S1: Collecting train running data through train sensors, performing train operation and track linkage processing based on the train running data, and generating train operation-track linkage data;

[0006] Step S2: performing real-time train positioning detection based on train running data and train operation-track linkage data to generate real-time train positioning data;

[0007] Step S3: Acquire train driving planning data; perform train driving schedule design based on the train driving planning data and the train driving real-time positioning data to generate train driving schedule data;

[0008] Step S4: analyzing the running status of each train carriage and the train traction force linkage based on the real-time train positioning data and the train running-track linkage data, and generating the running status of each train carriage-train traction force linkage data;

[0009] Step S5: Set the train adaptive asynchronous control instructions based on the operating status of each train car - train traction linkage data and train driving schedule data, generate train adaptive asynchronous control instruction data, and execute the train adaptive asynchronous control operation based on the train adaptive asynchronous control instruction data.

[0010] Furthermore, step S1 includes the following steps:

[0011] Step S11: collecting train running data through train sensors, performing train operation characteristic analysis based on the train running data, and generating train operation characteristic data;

[0012] Step S12: performing real-time train running track detection according to the train running characteristic data to generate real-time train running track data;

[0013] Step S13: analyzing train track structure change information based on the real-time train running track data to generate train track structure change information data;

[0014] Step S14: Perform train operation and track linkage processing based on the train operation characteristic data and the train track structure change information data to generate train operation-track linkage data.

[0015] Furthermore, step S2 includes the following steps:

[0016] Step S21: performing train-tower base station communication and train speed detection according to the train driving data, and generating train-tower base station communication data and train speed data respectively;

[0017] Step S22: monitoring train running environment changes based on train operation-track linkage data, and generating train running environment change data;

[0018] Step S23: performing train travel and tunnel travel positioning analysis based on train-tower base station communication data, train speed data, and train travel environment change data to generate train travel-tunnel travel positioning data;

[0019] Step S24: Perform real-time train positioning detection based on the train-tower base station communication data, train speed data, and train travel-tunnel travel positioning data to generate real-time train positioning data.

[0020] Furthermore, step S23 includes the following steps:

[0021] Step S231: Analyze the tunnel sequence and time difference of each train carriage according to the train running environment change data to generate tunnel sequence-time difference data of each train carriage;

[0022] Step S232: Based on the tunnel sequence-time difference data of each train car and the train-tower base station communication data, communication switching and recovery detection of each train car is performed to generate communication switching-recovery data of each train car;

[0023] Step S233: Calculating the train positioning delay based on the communication switching-restoration data of each train carriage and the train speed data to generate train positioning delay data;

[0024] Step S234: Perform train travel and tunnel travel positioning analysis based on the train travel positioning delay data and the train speed data to generate train travel-tunnel travel positioning data.

[0025] Furthermore, step S3 includes the following steps:

[0026] Step S31: obtaining train driving planning data, and dividing the train track occupation interval according to the train driving planning data to generate train track occupation interval data;

[0027] Step S32: Analyze each train intersection track section based on the train track occupation section data and the train driving real-time positioning data to generate each train intersection track section data;

[0028] Step S33: performing train freight load configuration analysis based on train driving data to generate train freight load configuration data;

[0029] Step S34: Based on the train freight load configuration data, the intersection track section data of each train, and the train track occupation interval data, the load adaptability matching of each train in the intersection track section is performed to generate the load adaptability data of each train in the intersection track section;

[0030] Step S35: Based on the train track occupancy section data, the load adaptability data of each train in the intersecting track section, and the intersecting track section data of each train, train schedule design is performed to generate train schedule data.

[0031] Furthermore, step S4 includes the following steps:

[0032] Step S41: performing train track curve and train track gradient detection based on the train real-time positioning data and the train operation-track linkage data, and generating train track curve data and train track gradient data respectively;

[0033] Step S42: performing a train curve track travel response dynamic analysis based on the train travel track curve data to generate train curve track travel response dynamic data;

[0034] Step S43: performing a train track gradient response dynamic analysis based on the train track gradient data to generate train track gradient response dynamic data;

[0035] Step S44: Based on the train curve track travel response power data and the train track gradient response power data, the running status of each train car and the train traction force linkage analysis are performed to generate the running status of each train car-train traction force linkage data.

[0036] Furthermore, step S42 includes the following steps:

[0037] Step S421: performing train contact axial center of gravity detection on the train operation-track linkage data based on the train track curve data to generate train contact axial center of gravity data;

[0038] Step S422: performing train track deflection displacement analysis based on the train contact axial center of gravity data to generate train track deflection displacement data;

[0039] Step S423: performing a train curve track contact area and resistance analysis based on the train track deflection displacement data to generate train curve track contact area-resistance data;

[0040] Step S424: Perform a train curve track travel response dynamic analysis based on the train curve track contact area-resistance data to generate train curve track travel response dynamic data.

[0041] Furthermore, step S43 includes the following steps:

[0042] Step S431: Detecting the traction spacing of each carriage on the train operation-track linkage data according to the train track slope data, and generating traction spacing data of each carriage;

[0043] Step S432: performing push-pull stress analysis on each carriage based on the traction spacing data of each carriage to generate push-pull stress data for each carriage;

[0044] Step S433: performing axial offset detection on each train based on the push-pull stress data of each carriage and the train operation-track linkage data to generate axial offset data of each train;

[0045] Step S434: Perform a train track gradient response dynamic analysis based on the axial offset data of each train and the push-pull stress data of each carriage to generate train track gradient response dynamic data.

[0046] Furthermore, step S44 includes the following steps:

[0047] Step S441: Analyze the longitudinal force coordination and speed difference of each train in the slope section based on the train track gradient response dynamic data, and generate longitudinal force coordination-speed difference data of each train in the slope section;

[0048] Step S442: Analyze the longitudinal traction deflection force of each train on the curved track based on the train curve track running response dynamic data to generate longitudinal traction deflection force data of each train on the curved track;

[0049] Step S443: Based on the longitudinal traction deflection force data of each train on the curved track and the longitudinal force coordination-speed difference data of each train in the slope section, the running status of each train car and the train traction force linkage analysis are performed to generate the running status of each train car-train traction force linkage data.

[0050] Furthermore, step S5 includes the following steps:

[0051] Step S51: Calculating the freight load configuration of each train car based on the train schedule data to generate freight load configuration data for each train car;

[0052] Step S52: Analyzing the traction power and braking of each train carriage based on the train carriage operation status-train traction force linkage data and the freight load configuration data of each train carriage, and generating traction power-braking data of each train carriage;

[0053] Step S53: Set the train adaptive asynchronous control instructions according to the traction power-braking data of each train car, generate train adaptive asynchronous control instruction data, and execute the train adaptive asynchronous control operation based on the train adaptive asynchronous control instruction data.

[0054] Beneficial effects of the present invention:

[0055] This paper proposes an artificial intelligence-based adaptive asynchronous control method for railway freight trains. Train sensors collect driving data and perform train operation and track linkage processing. The sensors can capture multi-dimensional data such as train speed, acceleration, carriage vibration, track flatness, curve radius, and slope in real time, ensuring data coverage of the entire train operation scenario. The linkage processing deeply correlates the train's operating status with the physical characteristics of the track. Real-time positioning detection based on train driving data and train operation and linkage data offers advantages in improved positioning accuracy and timely dynamic response. Combining train driving data with train operation and linkage data enables multi-source data cross-validation, reducing the impact of factors such as occlusion and signal delay. Dynamic updates of real-time positioning ensure instant capture of sudden train position changes. Train operation planning data is acquired and combined with real-time positioning data for train scheduling design, optimizing the dynamic adaptability of train scheduling and the efficiency of resource allocation. Scheduling design, combined with real-time positioning data, accurately matches deviations between actual train positions and planned paths, avoiding track resource conflicts and improving the overall transportation efficiency of the railway network. Based on real-time train positioning data, train operation, and linkage data, the system analyzes the linkage between car operating status and traction. This allows for refined management of train operating status and optimized configuration of traction output. This linkage analysis will clearly identify cars requiring higher traction support, while lightly loaded cars or cars on flat roads can appropriately reduce traction output, thereby reducing uneven pulling forces between cars caused by uneven traction distribution. Based on train operation and linkage data, as well as scheduling data, adaptive asynchronous control commands are set and control operations are executed. Differentiated control strategies can be developed for each car based on its operating status, traction requirements, and scheduling requirements. Asynchronous control enables smoother train operation in complex track environments.

[0056] The present invention provides an artificial intelligence-based adaptive asynchronous control method for railway freight trains. By locating the train's running position and tunnel communication blind spots in real time, and at the same time detecting and analyzing the changes in the curve radius and slope fluctuations of the running tracks of each train car, the method can monitor in real time the problems of uneven distribution of car states, delayed traction response, and accumulated positioning errors during train operation, thereby formulating adaptive asynchronous adjustment control instructions for the train, solving the problems of unbalanced traction coordination of each train car, excessive fluctuations in car stress, and uncontrolled driving rhythm, as well as coordinating and scheduling the use of tracks of multiple trains in intersecting track sections, avoiding scheduling conflicts among trains in intersecting track sections, thereby improving the transportation efficiency of freight trains and the utilization rate of track resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a schematic flow chart of the steps of an adaptive asynchronous control method for railway freight trains based on artificial intelligence according to the present invention;

[0058] Figure 2 for Figure 1 Detailed implementation steps of step S1 in FIG.

[0059] Figure 3 for Figure 1 Detailed implementation steps of step S2 in FIG.

[0060] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0061] The following is a clear and complete description of the technical method of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making any creative work are within the scope of protection of the present invention.

[0062] In addition, the accompanying drawings are merely schematic illustrations of the present invention and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor and / or microcontroller approaches.

[0063] It should be understood that although the terms "first," "second," and the like may be used herein to describe various elements, these elements should not be limited by these terms. These terms are used solely to distinguish one element from another. For example, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element, without departing from the scope of the exemplary embodiments. The term "and / or" as used herein includes any and all combinations of one or more of the listed associated items.

[0064] To achieve this, please refer to Figures 1 to 3 The present invention provides an artificial intelligence-based adaptive asynchronous control method for railway freight trains, comprising the following steps:

[0065] Step S1: Collecting train running data through train sensors, performing train operation and track linkage processing based on the train running data, and generating train operation-track linkage data;

[0066] Step S2: performing real-time train positioning detection based on train running data and train operation-track linkage data to generate real-time train positioning data;

[0067] Step S3: Acquire train driving planning data; perform train driving schedule design based on the train driving planning data and the train driving real-time positioning data to generate train driving schedule data;

[0068] Step S4: analyzing the running status of each train carriage and the train traction force linkage based on the real-time train positioning data and the train running-track linkage data, and generating the running status of each train carriage-train traction force linkage data;

[0069] Step S5: Set the train adaptive asynchronous control instructions based on the operating status of each train car - train traction linkage data and train driving schedule data, generate train adaptive asynchronous control instruction data, and execute the train adaptive asynchronous control operation based on the train adaptive asynchronous control instruction data.

[0070] In the embodiment of the present invention, please refer to Figure 1 FIG. 1 is a flow chart of steps of an adaptive asynchronous control method for railway freight trains based on artificial intelligence according to the present invention. In this example, the adaptive asynchronous control method for railway freight trains based on artificial intelligence includes the following steps:

[0071] Step S1: Collecting train running data through train sensors, performing train operation and track linkage processing based on the train running data, and generating train operation-track linkage data;

[0072] In an embodiment of the present invention, millimeter-wave radars, infrared temperature sensors, acceleration sensors installed at the front of the train, the bottom of the carriage and the wheelset, and optical fiber strain gauges and displacement sensors arranged along the track are used to collect data. The millimeter-wave radar captures the distance and relative speed between the train and the obstacle in front every second, the infrared temperature sensor monitors the friction temperature of the wheelset in real time, the acceleration sensor records the vertical and lateral vibration acceleration of the carriage, the optical fiber strain gauge transmits back the strain value of the track fastener every second, and the displacement sensor measures the track settlement. The collected speed, temperature, vibration, strain, and settlement data are input into a convolutional neural network model based on deep learning. The model is trained with multiple sets of historical data and can identify different track states (such as a curve radius of 300 meters and a slope of 2‰) and train operation parameters (such as a speed of 80km / h and an acceleration of 0.5m / s 2 ) and outputs linkage data including track flatness grade (1-5), wheel-rail friction coefficient (0.2-0.6), and train dynamic response coefficient.

[0073] Step S2: performing real-time train positioning detection based on train running data and train operation-track linkage data to generate real-time train positioning data;

[0074] In an embodiment of the present invention, the track curve parameters and slope values ​​in the linkage data are used in combination with the train GPS positioning data and the wheel pulse counter data for fusion calculation. The multi-source data is processed by the Kalman filter algorithm. The algorithm sets the process noise covariance matrix Q to 0.01I and the measurement noise covariance matrix R to 0.1I, and updates the positioning result every 0.2 seconds. Specifically, when the train travels to the track curve section, the curve radius value (such as 500 meters) in the linkage data is called to correct the drift of GPS due to signal obstruction. At the same time, the pulse count is compensated according to the wheel wear (the preset standard wheel diameter is 1.2 meters, and the actual measurement is 1.18 meters). The precise position of the train in the track mileage coordinate system is calculated (error ≤ 0.5 meters), and real-time positioning data including longitude and latitude, track section number (such as K123+450), and distance from the front switch (such as 300 meters) are generated. For example, when the GPS signal is lost in the tunnel, the positioning accuracy can still be maintained by the wheel diameter count and track slope data.

[0075] Step S3: Acquire train driving planning data; perform train driving schedule design based on the train driving planning data and the train driving real-time positioning data to generate train driving schedule data;

[0076] In an embodiment of the present invention, train operation planning data is obtained from the railway dispatching center database, including the planned arrival time of each station (such as arriving at station A at 10:00), the speed limit of the interval operation (such as 100km / h in a certain section), and the meeting avoidance point (such as K150). The real-time positioning data of the train (such as the current position K110+200, the actual time 9:40) and the planning data are input into the scheduling model based on the genetic algorithm. The model aims to minimize the total delay time, sets the population size to 50, the crossover probability to 0.8, and the mutation probability to 0.05, and calculates through multiple iterations. For example, when the real-time positioning shows that the train arrives at the K110 section 5 minutes later than planned, the model automatically adjusts the operating parameters of the subsequent sections, increases the target speed to 110km / h in the K120-K140 section with a speed limit of 120km / h, and compresses the stop time at station B by 3 minutes, and generates scheduling data including the expected arrival time of each section, speed curve, and meeting adjustment plan to ensure that the time deviation of the final arrival at the terminal is controlled within a certain range.

[0077] Step S4: analyzing the running status of each train carriage and the train traction force linkage based on the real-time train positioning data and the train running-track linkage data, and generating the running status of each train carriage-train traction force linkage data;

[0078] In the embodiment of the present invention, the current track condition of the train (such as 1.5‰ uphill, straight section) is determined based on the real-time positioning data, and the track resistance coefficient (such as 0.005) in the train operation and linkage data is combined to analyze the data of the pressure sensor (measuring load), speed sensor (measuring car speed), and current sensor (measuring traction motor current) installed in each car. A multivariate linear regression model is used, with car load (50-80 tons), track slope (-2‰ to 3‰), and running speed (60-120km / h) as independent variables and traction force (100-500kN) as dependent variable. The regression equation (R 2 =0.95). For example, when positioning data indicates the train is on a 2‰ uphill slope and the pressure sensor in car 3 indicates a 75-ton load, the model calculates that car requires 320kN of traction, while car 5, with a 50-ton load, requires only 210kN. The model then generates linkage data including the traction requirement for each car and the load-traction matching degree (0-100%).

[0079] Step S5: Set the train adaptive asynchronous control instructions based on the operating status of each train car - train traction linkage data and train driving schedule data, generate train adaptive asynchronous control instruction data, and execute the train adaptive asynchronous control operation based on the train adaptive asynchronous control instruction data.

[0080] In this embodiment of the present invention, train operation and linkage data (e.g., car 3 requires 320kN with a 92% matching degree) and scheduling data (e.g., the K130-K150 section needs to accelerate to 110km / h) are input into a fuzzy control model. This model includes three input variables (traction force demand deviation, scheduling time deviation, and car state coefficient) and two output variables (traction motor voltage adjustment and current adjustment). This model is calculated using a membership function (triangular distribution) and 49 control rules. For example, if car 3's actual traction force output is 300kN (with a deviation of -20kN) and the schedule requires that the section be advanced by 2 minutes, the model outputs control instructions for a 5% voltage increase and an 8% current increase. Simultaneously, for car 5, which has a lighter load, the model outputs instructions for a 3% voltage reduction and a 2% current reduction, thus implementing differentiated control. Control instructions are sent to the traction inverters in each carriage via the train Ethernet or the train's own local area network. The inverters adjust the IGBT switching frequency (1-10kHz) in real time according to the instructions to achieve asynchronous regulation of traction force.

[0081] Furthermore, step S1 includes the following steps:

[0082] Step S11: collecting train running data through train sensors, performing train operation characteristic analysis based on the train running data, and generating train operation characteristic data;

[0083] Step S12: performing real-time train running track detection according to the train running characteristic data to generate real-time train running track data;

[0084] Step S13: analyzing train track structure change information based on the real-time train running track data to generate train track structure change information data;

[0085] Step S14: Perform train operation and track linkage processing based on the train operation characteristic data and the train track structure change information data to generate train operation-track linkage data.

[0086] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 Detailed step flow diagram of step S1 in FIG. 1 , in this embodiment, step S1 includes the following steps:

[0087] Step S11: collecting train running data through train sensors, performing train operation characteristic analysis based on the train running data, and generating train operation characteristic data;

[0088] In an embodiment of the present invention, a laser Doppler velocimeter is installed on the top of the train head, a three-axis acceleration sensor is installed on the carriage bogie, and a temperature sensor is installed on the wheel set axle box. These sensors collect the real-time speed, longitudinal / lateral / vertical acceleration, and wheel set temperature data of the train. The collected continuous 60-minute data is divided into a time window of every 10 seconds and input into a classification model based on random forest. The model is trained with multiple sets of operation data under different loads (3000-5000 tons) and different track conditions, and can identify the train at a uniform speed (80km / h), acceleration (0.3m / s 2 ), deceleration (-0.4m / s 2 ) state, and outputs operational characteristic data including the train start-up acceleration curve (0-60km / h acceleration time: 120 seconds), braking deceleration curve (80-0km / h braking distance: 800 meters), wheelset temperature change rate (temperature rise of 5°C per hour), and vibration frequency distribution (mainly concentrated in the range of 5-15Hz). For example, when it is detected that the lateral acceleration exceeds 0.6g and lasts for 10 seconds when the train is accelerating, the model automatically marks it as an abnormal operational characteristic.

[0089] Step S12: performing real-time train running track detection according to the train running characteristic data to generate real-time train running track data;

[0090] In this embodiment of the present invention, vibration frequency and wheelset temperature change rates, derived from train operating characteristic data, are processed in conjunction with track image data collected by a track flaw detection radar and a high-definition linear array camera installed on the train's underside. The vibration frequency data is decomposed using a wavelet transform algorithm to extract vibration energy values ​​in the 4-8 Hz frequency band, which corresponds to loose trackbed characteristics. Simultaneously, the track images are processed using a machine vision Canny edge detection algorithm to identify the geometry of rail fasteners. The vibration energy value and fastener morphological parameters are input into the support vector machine model. The model sets the radial basis kernel function (parameter σ=0.8) and is trained through multiple groups of track status samples. It can output real-time running track data including track bed density (70%-95%), fastener integrity rate (85%-100%), rail wear (0-3mm), and gauge deviation (-2mm to +2mm). For example, when the vibration energy value is detected to be greater than 80dB and the fastener integrity rate is less than 90%, the model determines it as a loose track bed area and marks the specific location (K210+350 to K210+450).

[0091] Step S13: analyzing train track structure change information based on the real-time train running track data to generate train track structure change information data;

[0092] In this embodiment of the present invention, real-time train track data is differentially calculated on a daily, weekly, and monthly basis to extract track parameter variations. A sliding window algorithm (with a 7-day window size and a 1-day step size) is used to process rail wear data, calculating daily wear increments (0.01-0.05 mm / day). A Fourier transform is used to convert gauge deviation data into the frequency domain to identify periodic variation components (with a period of 30 days ± 5 days). A trendline fit (using the least squares method) is used to analyze the monthly rate of change in track bed density (-2% to 1% / month). These changes are input into an LSTM-based time series prediction model, which consists of three hidden layers (64 neurons per layer). After being trained on one year's track change data, the model can output track structure change information data, including rail wear trends (estimated wear of 0.8mm in the next 30 days), gauge deviation periodicity (increasing deviation from the 5th to the 10th of each month), trackbed settlement rate (0.5mm / month), and fastener loosening propagation rate (expanding 10 meters per day). For example, if the trackbed settlement rate in a certain section is calculated to exceed 0.8mm / month for three consecutive months, the model automatically generates track structure warning information and marks the starting mileage (K180+200) and affected range (200 meters) of the changed area.

[0093] Step S14: Perform train operation and track linkage processing based on the train operation characteristic data and the train track structure change information data to generate train operation-track linkage data.

[0094] In this embodiment of the present invention, a correlation analysis was performed between the braking distance and acceleration performance parameters in the train operation characteristic data and the rail wear and roadbed settlement rate in the track structure change information data. The Pearson correlation coefficient (ranging from -1 to 1) was used to determine the correlation between braking distance and rail wear (r=0.75), and the correlation between acceleration performance and roadbed settlement rate (r=-0.68). The correlation parameters are input into the graph neural network model, which contains 128 node feature dimensions and edge weights based on parameter influence settings (rail wear accounts for 40% and roadbed settlement accounts for 30%). After training with 50,000 sets of linkage samples, it can output operation-track linkage data including the correlation matrix between the train running resistance coefficient (0.012-0.018) and the track structure changes, the wheel-rail contact stress distribution (200-300MPa) changing with the wear amount, and the correspondence between the train dynamic response lag time (0.5-2 seconds) and the roadbed settlement. For example, when the rail wear is detected to reach 2.5mm and the running resistance coefficient rises to 0.017, the model outputs a linkage suggestion that the train needs to reduce the maximum speed by 5km / h, and marks the track section corresponding to the suggestion (K230+100 to K235+000).

[0095] Furthermore, step S2 includes the following steps:

[0096] Step S21: performing train-tower base station communication and train speed detection according to the train driving data, and generating train-tower base station communication data and train speed data respectively;

[0097] Step S22: monitoring train running environment changes based on train operation-track linkage data, and generating train running environment change data;

[0098] Step S23: performing train travel and tunnel travel positioning analysis based on train-tower base station communication data, train speed data, and train travel environment change data to generate train travel-tunnel travel positioning data;

[0099] Step S24: Perform real-time train positioning detection based on the train-tower base station communication data, train speed data, and train travel-tunnel travel positioning data to generate real-time train positioning data.

[0100] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Detailed step flow diagram of step S2 in the embodiment, step S2 includes the following steps:

[0101] Step S21: performing train-tower base station communication and train speed detection according to the train driving data, and generating train-tower base station communication data and train speed data respectively;

[0102] In an embodiment of the present invention, a 5G communication module installed on the train roof establishes a communication connection with a tower base station located every 5 kilometers along the line. The train identification code, current longitude and latitude, number of carriages, and other information are sent every 2 seconds. At the same time, data such as the track occupancy status of the 3 kilometers ahead and the signal display status (red light / green light / yellow light) are received from the tower base station. Train-tower base station communication data including communication signal strength (-50 to -100dBm), data transmission success rate (99.9%), and information exchange frequency are generated. At the same time, a Hall effect speed sensor is installed on the axle end of the train wheelset, outputting a pulse signal every 0.1 seconds. The instantaneous speed of the train (0-120km / h) is converted by a counter and calculated with the wheel diameter (1.2 meters ± 0.005 meters) compensation. Train speed data including speed change curve and acceleration value is generated. For example, when the communication signal strength is lower than -90dBm and lasts for 5 seconds, the train speed at this time is 85km / h.

[0103] Step S22: monitoring train running environment changes based on train operation-track linkage data, and generating train running environment change data;

[0104] In the embodiment of the present invention, the track flatness grade and wheel-rail friction coefficient in the train operation-track linkage data are called and processed in combination with the data collected by the meteorological sensors and infrared thermal imagers installed on both sides of the train. Through the threshold judgment method, when the temperature is below 0°C and the humidity is above 85%, it is marked as an icing-prone environment; at this time, the reference value of the track adhesion coefficient (0.3) needs to be multiplied by the correction factor of 0.8; the real-time data of the wind sensor is combined with the track direction (30°-60° east longitude), and the impact force is calculated by the formula F = 0.5×ρ×v 2 ×S×sinθ(where ρ is the air density 1.2kg / m 3 , v is wind speed, S is the lateral area of ​​the train, and θ is the angle between wind direction and track. The environmental parameters and linkage data are input into a Bayesian network model (containing five layers of nodes, with prior probabilities set based on multiple sets of historical data). The model then outputs data on train operating environment changes, including environmental risk levels (1-5), track adhesion coefficient correction values ​​(0.8-1.2 times the baseline value), and train running resistance increments (50-300N / t). For example, when a wind speed of 15m / s is detected and the track flatness level is Level 4, the model determines the environmental risk level to be Level 3 and marks the affected interval (K320+100 to K320+800).

[0105] Step S23: performing train travel and tunnel travel positioning analysis based on train-tower base station communication data, train speed data, and train travel environment change data to generate train travel-tunnel travel positioning data;

[0106] In this embodiment of the present invention, train and tunnel positioning analysis is performed based on signal status and track occupancy information from train-tower base station communication data, instantaneous speed and acceleration from train speed data, and environmental risk levels from train travel environment change data. In non-tunnel sections, base station triangulation is used to calculate the train's position by combining the signal reception times of three adjacent tower base stations with the known base station coordinates (latitude and longitude). When a train enters a tunnel (500-3000 meters in length), the RFID tags (storing tunnel mileage information) installed every 100 meters on the tunnel wall and the RFID reader on the train head are activated to read the tag information to obtain the relative position in the tunnel. At the same time, combined with the mileage accumulation in the train speed data (error ≤ 0.5 meters per 100 meters), the data is fused through the weighted average method (base station positioning weight 0.7, RFID positioning weight 0.3) to generate train travel-tunnel travel positioning data including the position switching points inside and outside the tunnel (accurate to 1 meter), the distance from the entrance / exit in the tunnel (0-3000 meters), and the positioning update frequency (1 time / 0.5 seconds). For example, when the train enters the 1000-meter-long K350 tunnel, at 300 meters from the entrance, the RFID positioning shows 298 meters, and the speed accumulation shows 302 meters. The fused positioning is 300 meters.

[0107] Step S24: Perform real-time train positioning detection based on the train-tower base station communication data, train speed data, and train travel-tunnel travel positioning data to generate real-time train positioning data.

[0108] In this embodiment of the present invention, train-tower base station communication data (including forward track status), train speed data (including instantaneous speed and acceleration), and train-tunnel positioning data (including precise position) are fed into an extended Kalman filter algorithm for fusion processing. The algorithm's state variables include position (x, y), velocity v, and acceleration a. The observed variables are base station positioning, RFID positioning, and speed sensor data. The process noise variance is set to 0.001, and the measurement noise variance is dynamically adjusted based on different scenarios (0.01 inside a tunnel and 0.005 outside a tunnel). Positioning results are updated every 0.2 seconds. Outliers are removed through residual analysis (residual ≤ 0.5 meters). Calibration is performed using track mileage stake numbers (one every 100 meters) from the train's real-time track data. This generates real-time train positioning data, including absolute position (error ≤ 0.3 meters), relative track mileage (accurate to 0.1 meters), and position confidence (95%-99.9%). For example, when the train reaches K400+500, the base station positioning shows K400+500.2, the RFID positioning (if in a tunnel) shows K400+499.8, and the speed accumulation shows K400+500.1. After algorithm fusion and calibration, the final positioning is K400+500.0, with a confidence level of 99.5%.

[0109] Furthermore, step S23 includes the following steps:

[0110] Step S231: Analyze the tunnel sequence and time difference of each train carriage according to the train running environment change data to generate tunnel sequence-time difference data of each train carriage;

[0111] In an embodiment of the present invention, the environmental risk level and tunnel section information in the train running environment change data are called, and the carriage spacing data collected by the laser ranging sensor installed at the head of each train carriage is analyzed. The laser ranging sensor outputs the distance value of the adjacent carriage every 0.5 seconds, and the total length from the front of the train to the nth carriage is calculated by cumulative calculation (each carriage is 14 meters long, and the total length of 15 carriages is 210 meters). When the train enters the tunnel entrance (based on the tunnel entrance mileage post), the time of the front of the train entering is recorded, and combined with the instantaneous speed in the train speed data (such as 80km / h, converted to 22.22m / s), the theoretical time for each carriage to enter the tunnel is calculated (the second carriage is 0.675 seconds later than the front, the third carriage is 0.675 seconds later than the second carriage, and so on). At the same time, the time difference is corrected according to the environmental risk level (such as level 3). When the wind speed is ≥10m / s, the time difference of each carriage is increased by 0.05 seconds. Generates sequence-time difference data including the time each carriage enters the tunnel (accurate to 0.01 seconds), the time it exits the tunnel, the time difference between carriages entering (0.675-0.725 seconds), and the total length adapted to the tunnel length. For example, when 15 carriages enter a 1000-meter tunnel, the entry time of the front of the train is 10:00:00.00, and the entry time of the 15th carriage is 10:00:10.12, which deviates by 0.01 seconds from the theoretically calculated 10:00:10.13.

[0112] Step S232: Based on the tunnel sequence-time difference data of each train car and the train-tower base station communication data, communication switching and recovery detection of each train car is performed to generate communication switching-recovery data of each train car;

[0113] In an embodiment of the present invention, the communication status of each carriage is monitored based on the sequence-time difference data of the carriage driving in the tunnel, combined with the signal strength and transmission success rate in the train-tower base station communication data. An independent 5G communication submodule (the same model as the main module of the front end) is installed on the top of each carriage to record the communication signal strength in real time. When the front end enters the tunnel, the time point when the second carriage enters the tunnel is predicted according to the sequence-time difference data, and the switching program of the communication submodule of the carriage is started 0.5 seconds in advance (switching from the tower base station communication to the emergency communication in the tunnel). By comparing the actual switching time with the predicted time (deviation ≤ 0.1 seconds), combined with the transmission success rate within 30 seconds after the switch (≥ 99.5%), it is judged whether the switch is successful; when the carriage exits the tunnel, the communication recovery program is started with a delay according to the sequence time difference based on the exit time of the front end, and the time when the signal strength returns to -80dBm after recovery is recorded (≤ 2 seconds). Generate communication switching-recovery data including the communication switching start time of each car, the switching success judgment result (success / failure), the recovery completion time, and the data packet loss rate during the switching (≤0.1%). For example, the 5th car started the switching at the predicted time 10:00:03.20, and the actual switching was completed at 10:00:03.28, with a deviation of 0.08 seconds. The transmission success rate after the switching was 99.8%, and the switching was judged to be successful.

[0114] Step S233: Calculating the train positioning delay based on the communication switching-restoration data of each train carriage and the train speed data to generate train positioning delay data;

[0115] In an embodiment of the present invention, the switching delay time (the time difference between the actual switching and the predicted time) and the recovery completion time in the communication switching-recovery data are used, combined with the acceleration and instantaneous speed in the train speed data to calculate the positioning delay. When the communication switching fails (such as the transmission success rate <99.5%), the failure duration is recorded (≤5 seconds), and the train travel distance (≤111.1 meters) during this period is calculated based on the speed data (such as 22.22m / s). This distance is the positioning lag caused by the communication interruption. For the carriages that successfully switch, the switching delay time (0.08 seconds) is multiplied by the speed to obtain the positioning delay distance (0.08×22.22≈1.78 meters). At the same time, when the carriage exits the tunnel and restores communication, the train travel distance (44.44 meters) within the recovery completion time (2 seconds) is calculated as the positioning delay in the recovery phase. The arithmetic average method is used to calculate the positioning delay of the entire train (the average value of the delay distance of each car), and the positioning delay data including the delay distance of the switching stage of each car (0-1.78 meters), the delay distance of the recovery stage (0-44.44 meters), the total delay distance (0-46.22 meters), and the delay duration are generated. For example, the switching delay of the 5th car is 1.78 meters, the recovery delay is 44.44 meters, and the total delay is 46.22 meters. There is no switching failure in the 8th car, the total delay is 1.80 meters, and the average delay of the entire train is 24.01 meters.

[0116] Step S234: Perform train travel and tunnel travel positioning analysis based on the train travel positioning delay data and the train speed data to generate train travel-tunnel travel positioning data.

[0117] In the embodiment of the present invention, the driving positioning delay data is called and combined with the mileage accumulation value in the train speed data to correct the train driving and tunnel driving positioning. The positioning delay distance is superimposed on the original positioning data (such as base station triangulation positioning or RFID positioning result) in the speed direction (forward direction is positive). When the delay distance is 46.22 meters, it is corrected to K350+346.22 based on the original positioning K350+300. At the same time, according to the delay duration (such as 5 seconds) and the speed change rate (acceleration 0.2m / s 2) and linear interpolation is used to calculate the delay correction per second (9.0 meters in the first second, 9.4 meters in the second second, and so on until the correction is complete after 5 seconds). Within the tunnel, the corrected positioning data is calibrated every 100 meters, combined with the absolute position read by the RFID tag (e.g., K350+500), to ensure a cumulative error of ≤1 meter. In non-tunnel sections, the base station triangulation results (error ≤3 meters) are fused with the corrected data, and a moving average method (window size of 5 data points) is used to smooth positioning fluctuations. Generate driving-tunnel driving positioning data including corrected tunnel positioning (accurate to 0.1 meter), non-tunnel positioning, delay correction curve (correction amount per second), and calibration point error (≤0.5 meter). For example, the original positioning of the train at K350+300 in the tunnel was displayed as K350+300 due to delay, and after correction it was K350+346.22. When it reaches the K350+500 RFID tag, the calibration shows the actual position as K350+500.3, with an error of 0.3 meters.

[0118] Furthermore, step S3 includes the following steps:

[0119] Step S31: obtaining train driving planning data, and dividing the train track occupation interval according to the train driving planning data to generate train track occupation interval data;

[0120] In an embodiment of the present invention, train driving planning data is retrieved from the railway dispatching center database, including information such as the planned route of each train (such as the K100-K500 interval), planned departure time (such as 7:00, 8:30), planned arrival time, and stopover sites (1 every 50 kilometers). The track occupation interval is divided in combination with the electronic mileage markers installed along the track (1 every 1 kilometer, storing accurate mileage information). An interval segmentation algorithm is used, with two adjacent stations as the basic unit (such as K100 station to K150 station as a unit), and then divided into sub-intervals every 10 kilometers. Each sub-interval is marked with a planned occupation time period (such as train A occupies K100-K110 from 8:00 to 8:10). At the same time, occupancy rules are set according to the track type (single-line / double-line). Only one train is allowed to occupy the same time period in the single-line interval, and the up and down lines of the double-line interval are occupied separately. Generate track occupation interval data including the interval number (such as Q100-1 represents K100-K110), planned occupied train number, occupation start time (accurate to minutes), occupation end time, interval length (10 kilometers), and track type. For example, train B plans to occupy the double-track interval Q150-2 (K150-K170) from 9:00 to 9:20, and there is no conflict with train C occupying the opposite interval Q150-3 from 9:10 to 9:30.

[0121] Step S32: Analyze each train intersection track section based on the train track occupation section data and the train driving real-time positioning data to generate each train intersection track section data;

[0122] In an embodiment of the present invention, based on the data on the occupied intervals of the train tracks, combined with the real-time positioning data of the trains (including the current mileage and speed of each train), the train intersection situation is analyzed. By establishing a train position-time matrix, with the horizontal axis being time (in minutes) and the vertical axis being track mileage, the occupied interval data of each train is converted into a rectangular area in the matrix (such as train A at 8:00-8:10 corresponds to K100-K110). The real-time positioning data is called, and the coordinates of the current position of each train in the matrix are marked (such as train D at K120 at 8:05). The collision detection algorithm is used to calculate the overlap of the rectangular areas of different trains. The area with an overlap greater than 0 is a potential intersection section. Further analysis of the potential section shows that when the speeds of the two trains are 80km / h and 90km / h respectively, and the distance is less than 5 kilometers, it is determined to be an intersection section that requires special attention. Generate intersection track section data including the intersection section mileage (such as K130-K135), estimated intersection time (such as 8:20), involved train numbers (2 trains), planned speeds at intersection (80km / h and 70km / h), and whether the track is an avoidance section (yes / no). For example, train numbers E and F are expected to intersect at 8:20 in the K130-K135 single-track section, which is equipped with an avoidance platform (500 meters long).

[0123] Step S33: performing train freight load configuration analysis based on train driving data to generate train freight load configuration data;

[0124] In an embodiment of the present invention, freight load configuration analysis is performed based on the real-time load data collected by the carriage load sensors in the train driving data (installed at the bottom of each carriage, with a measurement range of 0-100 tons and an accuracy of ±0.5 tons). The load sensor of each carriage uploads data every 10 minutes (such as 50 tons for the first carriage and 60 tons for the second carriage), and the total load of the train is obtained by accumulation (the total load of 15 carriages is 800 tons ±5 tons). Combined with the cargo type information (such as coal, steel), the uniformity of the load distribution of each carriage is calculated (standard deviation ≤ 5 tons). A uniformity of <3 tons is excellent, 3-5 tons is good, and >5 tons is poor. At the same time, based on the track slope data, the impact of the load on the climbing ability is analyzed. When the slope is >3‰, the maximum load limit of each carriage is 60 tons. Generate freight load configuration data including car number, single-car load (tons), total load (tons), load uniformity grade, cargo type adaptability (such as steel adaptability 90%), and load compliance under slope restrictions (compliant / non-compliant). For example, the load of car No. 8 is 65 tons, which is judged as non-compliant in the 3‰ slope section and needs to be marked for adjustment.

[0125] Step S34: Based on the train freight load configuration data, the intersection track section data of each train, and the train track occupation interval data, the load adaptability matching of each train in the intersection track section is performed to generate the load adaptability data of each train in the intersection track section;

[0126] In an embodiment of the present invention, freight load configuration data (including the total load and uniformity of each train), intersection track section data (including intersection section slope and length) and train track occupation interval data (including interval speed limit) are called to perform load adaptability matching. Using the load-slope matching algorithm, when the intersection section slope is 2‰, trains with a total load of ≤1000 tons can maintain 80km / h to pass, and 1000-1200 tons need to be reduced to 70km / h. Combined with the load uniformity of each train, trains with good uniformity can increase their speed by 5km / h when intersecting on a curve (curve radius of 300 meters), and trains with poor uniformity need to reduce their speed by 5km / h. At the same time, based on the time overlap of the track occupation interval, the safe distance (≥500 meters) between the two trains in the intersection section is calculated, and heavily loaded trains need to slow down 10 seconds in advance. Generate load adaptability data including the recommended speed of each train in the intersection section (60-80km / h), deceleration starting mileage (such as K129.5), acceleration ending mileage (such as K135.5), load and section adaptation level (excellent / good / poor), and safety distance margin (500-800 meters). For example, the total load of train No. G is 900 tons (uniformity is excellent), the recommended speed in the 2‰ slope intersection section is 75km / h, and the safety distance margin is 600 meters.

[0127] Step S35: Based on the train track occupancy section data, the load adaptability data of each train in the intersecting track section, and the intersecting track section data of each train, train schedule design is performed to generate train schedule data.

[0128] In an embodiment of the present invention, a train schedule is designed based on the data of the intervals occupied by the train tracks, the load adaptability data of each train in the intersection track section, and the data of each train intersection track section. A genetic algorithm is used to optimize the schedule, with minimizing the total delay time (target value ≤ 5 minutes) as the fitness function, the population size is set to 100, the crossover probability is 0.7, the mutation probability is 0.05, and 50 iterations are performed. The recommended speed and safety distance of the intersection section are converted into constraint conditions. When two trains intersect on a single line, the heavily loaded train (such as 1,200 tons) is given priority to pass first, and the other train stops at the avoidance station (stop time 5 minutes). Combined with the real-time positioning data of the train, the speed of the subsequent section is adjusted for trains that deviate from the plan (such as a delay of 2 minutes) (increase by 5 km / h within the speed limit). Generate driving schedule data including adjusted departure time (accurate to the minute), speed in each section (60-90 km / h), intersection avoidance plan (who passes first), estimated arrival time (deviation ≤ 1 minute), and scheduling adjustment basis (such as priority for heavy loads). For example, due to a 2-minute delay, train H will increase its speed from 80 km / h to 85 km / h in the K200-K250 section to ensure arrival at K250 station at 10:00 as planned, and the intersection time with the subsequent train will be adjusted to 10:10.

[0129] Furthermore, step S4 includes the following steps:

[0130] Step S41: performing train track curve and train track gradient detection based on the train real-time positioning data and the train operation-track linkage data, and generating train track curve data and train track gradient data respectively;

[0131] In an embodiment of the present invention, track curve and slope detection is performed based on real-time train positioning data (including precise mileage and position coordinates) and train operation-track linkage data (including track structural parameters). A laser profile scanner is installed at the bottom of the train head to obtain track cross-sectional profile data every second. Combined with the mileage information (such as K500+200) in the real-time positioning data, the track curve radius is calculated using a curve fitting algorithm (using the least squares method). When the contour within a continuous range of 50 meters meets the arc characteristics, it is determined to be a curve segment, and the curve radius (300-1000 meters), curve length (50-500 meters), and curve start / end mileage of the curve are output. At the same time, the track inclination angle is collected using the inclination sensor installed on the train bogie (measuring range -15° to 15°, accuracy ±0.01°). Combined with the altitude information in the real-time positioning data (error ±0.5 meters), the slope value is calculated using the slope calculation formula, and a slope data (-5‰ to 5‰) is output every 10 meters to generate the driving track slope data including the starting / ending mileage of the slope, slope value, slope length (100-1000 meters), and slope type (uphill / downhill). For example, in the section from K510+300 to K511+300, a curve radius of 500 meters is detected, and there is a 2‰ uphill slope, which generates curve data and slope data respectively.

[0132] Step S42: performing a train curve track travel response dynamic analysis based on the train travel track curve data to generate train curve track travel response dynamic data;

[0133] In an embodiment of the present invention, a dynamic analysis of the curved track travel response is performed based on the train track curve data (including curve radius and length). The curve radius of 300 meters and the curve length of 200 meters in the curve data are called, and combined with the speed (80km / h) and car wheelbase (15 meters) in the train driving data, a multi-body dynamics simulation model (including 15 car rigid bodies and a spring damping system) is used for analysis. In the model, the wheel-rail contact force is calculated using Hertz contact theory. The superelevation value of the track curve section is set (150mm, corresponding to a 300-meter radius). The train speed is input to calculate the centrifugal force (the total mass of the train is 800 tons), and the lateral impact force (50-150kN) is obtained. At the same time, the actual lateral vibration data is collected by the lateral acceleration sensors (range ±2g, accuracy ±0.01g) installed on both sides of the car, and compared with the simulation results to correct the dynamic response parameters. Generate curved track travel response dynamic data including lateral acceleration peak (0.5-1.2g), wheel-rail lateral force (80-200kN), centripetal force required for curve passing (100-300kN), car roll angle (1-3°), and curve travel dynamic safety factor (≥1.2). For example, when a train passes a 300-meter radius curve at 80km / h, the lateral acceleration peak is 0.8g, the wheel-rail lateral force is 150kN, and the safety factor is 1.5, which meets safety standards.

[0134] Step S43: performing a train track gradient response dynamic analysis based on the train track gradient data to generate train track gradient response dynamic data;

[0135] In an embodiment of the present invention, a dynamic analysis of the track slope response is performed based on the slope data of the train track (including slope value and length). For the 2‰ uphill slope and 1000-meter length in the slope data, combined with the total train load of 800 tons and the current speed of 80 km / h, dynamic analysis software is used to calculate the traction required for slope travel (the calculation formula is F = mgsinθ + fmgcosθ, θ is the slope angle, f is the rolling friction coefficient of 0.002, m is the total train load, and g is the acceleration of gravity). The theoretical traction force (200-500 kN) is obtained. The actual output torque is collected by the torque sensor installed on the traction motor (range 0-10000 N·m, accuracy ±10 N·m), converted into actual traction force (220 kN), and compared with the theoretical value to calculate the deviation (10%). At the same time, the speed change of the train on the slope is analyzed. When the actual traction force is greater than the theoretical value, the speed increase rate is 0.1 m / s. 2 , otherwise the descent rate is 0.05m / s 2 Generate the theoretical traction required for the slope, actual traction, traction deviation rate (5%-15%), speed change rate (±0.1m / s 2), the track slope response power data of the slope driving power reserve coefficient (actual traction / theoretical traction, ≥1.1), for example, in a 2‰ uphill section, the theoretical traction force is 200kN, the actual output is 220kN, the power reserve coefficient is 1.1, and the speed is 0.1m / s 2 Lift to meet the requirements of slope driving.

[0136] Step S44: Based on the train curve track travel response power data and the train track gradient response power data, the running status of each train car and the train traction force linkage analysis are performed to generate the running status of each train car-train traction force linkage data.

[0137] In the embodiment of the present invention, based on the train curve track running response dynamic data (including lateral force and centripetal force) and the train track slope response dynamic data (including traction force and speed change rate), the carriage running state and traction force linkage analysis are performed. The lateral force of 150kN and centripetal force of 200kN in the curve dynamic data are called, and the actual traction force of 220kN and speed change rate of 0.1m / s in the slope dynamic data are called. 2 Calculations are performed using a traction force distribution algorithm, combining data from each car's load sensor (50 tons for car 1, 60 tons for car 2, and 55 tons for car 15). The algorithm allocates base traction force based on the car's load ratio (16% for a car with a 60-ton load, 13% for a car with a 50-ton load). A lateral correction factor is added based on curve dynamics data (5% increase for cars on the inside of a curve, 5% decrease for cars on the outside), and a slope correction factor is added based on slope data (10% increase for all cars on uphill slopes). Current sensors (range 0-500A, accuracy ±1A) installed in each car's traction motor collect actual current and convert it into actual traction force for each car (15-25kN). This is then compared with the calculated value, and secondary adjustments are made if the deviation exceeds 5%. Generate carriage operation status-traction linkage data, including the theoretical traction of each carriage, actual traction, traction distribution ratio (5%-18%), traction and lateral force matching (≥80%), and speed synchronization rate of each carriage (≥95%). For example, the second carriage with a load of 60 tons has a theoretical traction of 24kN and an actual output of 23.5kN in the curve + uphill section, with a deviation of 2% and a matching degree of 92%, which meets the linkage requirements.

[0138] Furthermore, step S42 includes the following steps:

[0139] Step S421: performing train contact axial center of gravity detection on the train operation-track linkage data based on the train track curve data to generate train contact axial center of gravity data;

[0140] In an embodiment of the present invention, based on the train track curve data (including a curve radius of 300 meters and a curve length of 200 meters), combined with the wheel-rail friction coefficient of 0.35 in the train operation-track linkage data, the train contact axial center of gravity is detected. Three displacement sensors are installed on the inner side of each pair of wheels on the train (measuring range -50 to 50mm, accuracy ±0.1mm), and the lateral distance between the wheelset and the inner side of the rail is collected respectively (130mm on the left and 128mm on the right), and the data is recorded every 0.05 seconds. Through the center of gravity calculation model, with the center of the wheelset as the origin, the lateral distance data is converted into the wheelset center of gravity offset (left offset +2mm, right offset -2mm), and the center of gravity offset trend under the action of centrifugal force is calculated in combination with the curve radius (the offset outside the curve increases by 5mm). At the same time, the wheel-rail contact point distribution in the linkage data is called to determine the reference axis of the contact axis (coinciding with the center line of the track), and the deviation value (±3mm) of the actual center of gravity and the reference axis is calculated. Generate contact axial center of gravity data including the three-dimensional coordinates of the wheelset center of gravity (X-axis along the track direction, Y-axis horizontally, and Z-axis vertically, accurate to 0.1mm), the center of gravity offset of each wheelset (-5 to +5mm), the axial center of gravity offset rate (0.2mm / s), and the maximum center of gravity offset of the curve segment (+6mm). For example, at the midpoint of a 300-meter radius curve, the coordinates of the center of gravity of the fifth pair of wheelsets are (K510+400,+3mm,0mm), which is +4mm offset from the straight segment.

[0141] Step S422: performing train track deflection displacement analysis based on the train contact axial center of gravity data to generate train track deflection displacement data;

[0142] In an embodiment of the present invention, track deflection displacement analysis is performed based on the axial center of gravity data of the train (including the center of gravity offset and the offset rate). A finite element analysis model is used to divide the track structure into 1-meter-long units. The lateral force of each wheel pair in the center of gravity offset data is input (calculated based on the offset, an offset of +5mm corresponds to a lateral force of 80kN) to simulate the elastic deformation of the rail under the action of the lateral force. Laser displacement sensors (measuring range 0-100mm, accuracy ±0.05mm) are installed on both sides of the track to collect the lateral displacement of the top of the rail in real time (rail displacement +3mm on the outside of the curve, -2mm on the inside), and output data every 0.1 seconds. By comparing the theoretical displacement values ​​of the finite element model (+3.2mm on the outside, -1.8mm on the inside) with the actual measured values, the displacement deviation rate (6.25%) is calculated, and the track displacement change trend (increase of 0.5mm per 10 meters) is inferred based on the center of gravity offset rate. The system generates track deflection displacement data including the rail lateral displacement value (-3 to +4 mm), displacement deviation rate (5%-8%), track elastic deformation coefficient (0.02 mm / kN), and curve section displacement distribution gradient (the outer side increases by 2 mm from the starting point to the midpoint of the curve). For example, at K510+350 of a 300-meter radius curve, the outer rail displacement is +3 mm, which deviates from the theoretical value by 0.2 mm and is within the allowable range.

[0143] Step S423: performing a train curve track contact area and resistance analysis based on the train track deflection displacement data to generate train curve track contact area-resistance data;

[0144] In the embodiment of the present invention, the contact area and resistance of the curved track are analyzed based on the deflection displacement data of the train track (including the lateral displacement of the rail and the deformation coefficient). A pressure sensitive film (accuracy 0.1N / mm) is attached to the surface of the wheelset. 2 ), the film compression marks were photographed with a high-speed camera (frame rate 1000fps), the major axis (80mm±5mm) and minor axis (30mm±2mm) of the wheel-rail contact area were identified, and the contact area (S0=a×b, where a is the major axis contact length and b is the minor axis contact length, 2400mm) was calculated. 2 ±200mm 2 Combined with the rail deformation in the track deflection displacement, when the lateral displacement is +3mm, the contact area is reduced by 10% (2160mm 2 At the same time, based on the contact area and the wheel-rail friction coefficient of 0.35, the rolling resistance is calculated using Coulomb's friction law (F = μN, N is the vertical load of the wheelset 80kN, μ is the wheel-rail friction coefficient of 0.35), and the resistance value of the curved section is obtained (28kN), which is 12% higher than that of the straight section (25kN). Generate a curve containing the wheel-rail contact area (2000-2600mm 2), contact area change rate (-15% to +5%), contact pressure per unit area (30-40N / mm 2 ), rolling resistance in curves (25-30kN), the relationship between the resistance increase and the curve radius (300m radius corresponds to an increase of 3kN), the curve track contact area-resistance data, for example, at a 300m radius curve, the contact area is 2200mm 2 The rolling resistance is 28kN, which is 3kN higher than that of the straight section, which is consistent with the theoretical calculation.

[0145] Step S424: Perform a train curve track travel response dynamic analysis based on the train curve track contact area-resistance data to generate train curve track travel response dynamic data.

[0146] In the embodiment of the present invention, based on the train curve track contact area-resistance data (including contact area and rolling resistance), the curve track driving response dynamic analysis is performed. The contact pressure per unit area in the contact area data is called 35N / mm 2 , combined with the elastic deformation coefficient in the track deflection displacement data, the power transmission efficiency of the wheel-rail contact (90%) is calculated. Using the dynamic response model, the rolling resistance of 28kN and the train speed of 80km / h are input to calculate the additional power required for curve driving (overcoming resistance + providing centripetal force, a total of 150kN). The longitudinal (0.1g), lateral (0.8g), and vertical (0.3g) accelerations during curve driving are collected using a three-dimensional accelerometer installed on the bottom of the car (range ±2g, accuracy ±0.01g). The accelerations are compared with the acceleration values ​​calculated by the model (longitudinal 0.09g, lateral 0.82g) to correct the dynamic response parameters. Generate curved track travel response dynamic data including total power demand for curved travel (150kN), wheel-rail contact power transmission efficiency (88%-92%), peak acceleration in each direction (0.8g in lateral direction), dynamic response lag time (0.1 second), and curve travel dynamic stability coefficient (≥1.2). For example, when a train passes through a 300-meter radius curve at 80km / h, the total power demand is 150kN, the lateral acceleration peak is 0.8g, and the stability coefficient is 1.4, which meets the dynamic response safety standards.

[0147] Furthermore, step S43 includes the following steps:

[0148] Step S431: Detecting the traction spacing of each carriage on the train operation-track linkage data according to the train track slope data, and generating traction spacing data of each carriage;

[0149] In an embodiment of the present invention, according to the train track gradient data (including 2‰ uphill, length 1000 meters), combined with the total length of the train in the train operation-track linkage data of 210 meters, the traction spacing of each carriage is detected. A laser ranging sensor (measuring range 0-50 meters, accuracy ± 0.05 meters) is installed at the head and tail connection of each carriage, and the distance between adjacent carriages is collected every 0.1 seconds (the standard spacing of the straight section is 1.5 meters ± 0.05 meters). When the train enters the uphill section, the spacing change is recorded in real time. The spacing between the first and second carriages increases to 1.55 meters, the spacing between the second and third sections increases to 1.53 meters, and so on. The spacing between the rear carriages changes the least (1.51 meters). At the same time, combined with the train speed of 80km / h in the linkage data, the spacing change rate (0.02 meters / second) is calculated, and the slope value is associated to establish a spacing-slope relationship model (for every 1‰ increase in slope, the spacing increases by an average of 0.03 meters). The traction spacing data of each carriage is generated, including the real-time spacing between each carriage (1.5-1.6 meters), the spacing change rate (0.01-0.03 meters / second), the maximum spacing in the uphill section (1.6 meters), the deviation of each spacing from the standard value (0.01-0.1 meters), and the spacing distribution gradient (gradually decreasing from the front to the rear of the train). For example, at the midpoint K510+800 of the 2‰ uphill section, the spacing between the 5th and 6th carriages is 1.54 meters, which deviates by 0.04 meters from the straight section, which is in line with the spacing change law in the slope section.

[0150] Step S432: performing push-pull stress analysis on each carriage based on the traction spacing data of each carriage to generate push-pull stress data for each carriage;

[0151] In this embodiment of the present invention, push-pull stress analysis of each car is performed based on the traction spacing data of each car (including real-time spacing and rate of change). A spring-damper mechanics model is used to simulate the car connection device. The change in spacing between each car in the spacing data is converted into the stretch of the connection device (a spacing of 1.55 meters corresponds to a stretch of 0.05 meters). The tensile force is calculated according to Hooke's law, resulting in a tensile force of 52 kN between the first car and the second car, and 48 kN between the second car and the third car. Simultaneously, strain gauge sensors (measuring range -2000 to 2000 με, accuracy ±5 με) are installed at the connection device to collect stress and strain values ​​in real time (a strain of 1000 με at the connection between the first and second cars corresponds to a stress of 50 kN). The calculated values ​​are compared with the measured values, and the deviation is controlled within 5%. Combined with the slope data, which shows a slope length of 1000 meters, the stress duration is analyzed. When the spacing is stable at 1.55 meters, the stress remains at 52 kN for 10 seconds. Generate push-pull stress data for each carriage, including the push-pull stress value (40-60kN) between each carriage, stress change rate (2kN / second), stress-to-spacing conversion coefficient (1 meter spacing corresponds to 1000kN), maximum stress value in the uphill section (60kN), and stress distribution trend (gradually decreasing from the front to the rear of the train). For example, in the 2‰ uphill section, the stress at the connection between the 1st and 2nd carriages is 52kN, which deviates from the calculated value by 2kN and is within the allowable error range. The stress distribution is consistent with the characteristics of the front of the train bearing the main traction force.

[0152] Step S433: performing axial offset detection on each train based on the push-pull stress data of each carriage and the train operation-track linkage data to generate axial offset data of each train;

[0153] In an embodiment of the present invention, based on the push-pull stress data (including stress value and rate of change) of each car, combined with the coordinates of the wheel-rail contact point in the train operation-track linkage data, the axial offset of each train is detected. A two-dimensional laser displacement sensor (measuring range -100 to 100mm, accuracy ±0.1mm) is installed at the bottom of each car, and the offset of the car relative to the center line of the track is collected along the longitudinal and transverse directions of the track (longitudinal offset ±0.5mm, transverse offset ±2mm). According to the stress difference in the push-pull stress data (the tension of the first section to the second section is 52kN, the tension of the second section to the third section is 48kN, and the stress difference is 4kN), the axial deflection angle (0.1°) of the car caused by uneven force is calculated and converted into a longitudinal offset (0.3mm for each car). The coordinates of the center line of the track in the linkage data are called to determine the axial offset baseline, and the measured offset is compared with the calculated offset. The deviation is controlled within 0.2mm. The axial offset data of each train is generated, including the longitudinal offset (-0.6 to +0.6mm), lateral offset (-2 to +2mm), axial deflection angle (0-0.2°), correlation between offset and stress difference (0.8), and maximum offset position (3rd car +0.6mm). For example, in the 2‰ uphill section, the 3rd car has a longitudinal offset of +0.5mm and a lateral offset of +1mm due to the front-rear stress difference of 5kN, which deviates from the calculated value by 0.1mm, which is in line with the force offset law.

[0154] Step S434: Perform a train track gradient response dynamic analysis based on the axial offset data of each train and the push-pull stress data of each carriage to generate train track gradient response dynamic data.

[0155] In this embodiment of the present invention, a dynamic analysis of the train track gradient response is performed based on the axial offset data (including offset and deflection angle) of each train and the push-pull stress data (including stress values) of each car. The longitudinal offset in the axial offset data is converted into additional resistance (10N is added for every millimeter of offset), and the lateral offset is converted into lateral friction resistance between the wheel and rail (5N is added for every millimeter of offset). The additional resistance generated by the offset of the third car is 5N + 5N = 10N. Combined with the total tension in the push-pull stress data (the sum of the stresses in each car is 800kN), the effective traction force (total tension minus additional resistance, 799.9kN) is calculated using the dynamic balance equation. Compared with the theoretical traction force of 200kN in step S43, the power utilization rate is calculated (799.9kN / 800kN = 99.99%). The actual output power is collected by the power sensor installed in the traction motor and converted to the actual power output (800kW), which deviates from the calculated value by 0.1%. Generate train track slope response dynamic data including effective traction in the slope section, total additional resistance (50-100N), power utilization rate (99.5%-100%), power distribution ratio of each car (15% at the front of the car, decreasing to 5% at the rear of the car), and slope response dynamic stability coefficient (effective traction / theoretical traction ≥ 0.95). For example, in a 2‰ uphill section, the total additional resistance is 80N, the effective traction is 799.92kN, the power utilization rate is 99.99%, and the stability coefficient is 1.0, which meets the slope section dynamic response requirements and the power distribution of each car conforms to the stress distribution law.

[0156] Furthermore, step S44 includes the following steps:

[0157] Step S441: Analyze the longitudinal force coordination and speed difference of each train in the slope section based on the train track gradient response dynamic data, and generate longitudinal force coordination-speed difference data of each train in the slope section;

[0158] In an embodiment of the present invention, the longitudinal force coordination and speed difference of the train in the slope section are analyzed based on the dynamic data of the train track gradient response (including effective traction and power distribution ratio). A torque sensor and a speed sensor are installed at the traction motor of each car of the train, and data is collected once per second. For the 2‰ uphill section, the deviation between the actual traction of each car and the theoretical value of the power distribution ratio is calculated. For example, the theoretical traction of the first car is 120kN, the actual traction is 118kN, and the deviation is -1.7%; the theoretical traction of the fifth car is 80kN, the actual traction is 82kN, and the deviation is +2.5%. The deviation of all cars is controlled within ±3%, and it is determined to be well coordinated. At the same time, the speed difference of adjacent cars is calculated. The speed of the first car is 80km / h, and the speed of the second car is 79.8km / h, with a difference of 0.2km / h. The maximum speed difference of the entire train is 0.3km / h. The force coordination coefficient formula (1-average of the absolute values ​​of the traction deviations of each car) is calculated to be 0.98, and the speed difference coefficient (1-maximum speed difference / average speed) is 0.996. The longitudinal force coordination-speed difference data is generated, including the traction deviation rate of each car (-3% to +3%), the speed difference between adjacent cars (0.1-0.3km / h), the longitudinal force coordination coefficient (≥0.95), the speed difference coefficient (≥0.99), and the force coordination stability of the slope section (no deviation for 5 minutes). For example, on a 2‰ uphill section, the force coordination coefficient of the entire train is 0.98 and the speed difference coefficient is 0.996, which meets the operating requirements of the slope section.

[0159] Step S442: Analyze the longitudinal traction deflection force of each train on the curved track based on the train curve track running response dynamic data to generate longitudinal traction deflection force data of each train on the curved track;

[0160] In the embodiment of the present invention, the longitudinal traction deflection force of the curved track train is analyzed based on the dynamic data of the train's curved track travel response (including lateral force and centripetal force). A three-dimensional force sensor (range -500 to 500kN, accuracy ±1kN) is installed at the wheelset of each carriage to collect the longitudinal, lateral and vertical forces between the wheel and rail during curve travel, and record them every 0.05 seconds. For a 300-meter radius curve, combined with the lateral acceleration of 0.8g in the dynamic data of the curved track travel response, the longitudinal traction deflection force is calculated (the longitudinal traction deflection force formula is F=F h ×tanθ,F h(where θ is the lateral force and the curve's turning angle (a 300-meter radius corresponds to a turning angle of 1.91°), the deflection force on the first car is 15 kN, on the fifth car 20 kN, and on the rear car 10 kN. Comparing the actual longitudinal forces measured by the 3D force sensor (14.8 kN for the first car and 20.2 kN for the fifth car) with the calculated values ​​reveals a deviation within ±2%. Furthermore, analysis of the relationship between deflection force and curve position reveals that the deflection force increases from the curve's starting point to the midpoint and decreases from the midpoint to the end point, with the maximum deflection force occurring at the midpoint. Generate curved track longitudinal traction deflection force data including the longitudinal traction deflection force value of each car (10-25kN), the ratio of deflection force to lateral force (0.03-0.05), the deviation rate between actual and calculated deflection force (-2% to +2%), the curve section deflection force distribution curve, and the maximum deflection force position (midpoint of the curve). For example, at the midpoint of a 300-meter radius curve, the longitudinal traction deflection force of the fifth car is 20kN, which deviates from the calculated value by 0.2kN, is within the allowable range, and is consistent with the characteristic that the force is greatest in the middle of the curve section.

[0161] Step S443: Based on the longitudinal traction deflection force data of each train on the curved track and the longitudinal force coordination-speed difference data of each train in the slope section, the running status of each train car and the train traction force linkage analysis are performed to generate the running status of each train car-train traction force linkage data.

[0162] In this embodiment of the present invention, a linkage analysis of train car operating status and traction is performed based on the longitudinal traction deflection force data of curved tracks and the longitudinal force coordination-speed difference data of ramp sections. The deflection force values ​​of each car in the deflection force data are called up, combined with the traction force deviation rate in the coordination data, and a dynamic traction force distribution model is used to modify the basic traction force (80kN for a 50-ton car and 96kN for a 60-ton car, based on the load distribution): For cars with deflection force greater than 15kN in curved sections, the traction force is increased by 5% (from 96kN to 100.8kN for the fifth section), and for cars with speed difference greater than 0.2km / h in ramp sections, a fine-tuning of ±2% is performed (from 88kN to 89.76kN for the second section). A traction force actuator feedback sensor (range 0-200kN, accuracy ±0.5kN) is installed in each car to collect the actual traction force after adjustment in real time and compare it with the model output value, with a deviation of ≤1kN. At the same time, analysis of the adjusted carriage operating status shows a 10% reduction in lateral vibration amplitude and a reduction in speed difference to within 0.1 km / h. The system generates carriage operating status-traction linkage data, including the adjusted traction force (80-120 kN) of each carriage, the traction force adjustment amplitude (-2% to +5%), the adjusted lateral vibration value (0.4-0.7 g), the adjusted speed difference (0.1-0.2 km / h), and the matching degree between traction and operating status (≥0.95). For example, on a 300-meter radius curve with a 2‰ uphill section, the fifth carriage has an adjusted traction force of 100.8 kN, a lateral vibration of 0.6 g, and a speed difference of 0.1 km / h from the previous carriage, resulting in a matching degree of 0.98, achieving precise linkage between traction and operating status.

[0163] Furthermore, step S5 includes the following steps:

[0164] Step S51: Calculating the freight load configuration of each train car based on the train schedule data to generate freight load configuration data for each train car;

[0165] In the embodiment of the present invention, the freight load configuration of each carriage is calculated based on the train schedule data (including the planned speed of each section and the arrival time). The train dispatching system imports the speed limit of the section (such as 90km / h in the K500-K600 section), slope information (2‰ uphill), and curve parameters (300m radius) in the schedule data, and combines the rated load of each carriage (60 tons) and the type of cargo (density 2.5 tons / m 3) and distributes cargo using a load-balancing algorithm. This algorithm targets a 5% deviation in axle weight between each car, adjusting loads based on power requirements within the schedule: The load on the first three cars on uphill sections is reduced by 5% (57 tons), while the load on cars in the middle of curved sections is reduced by 3% (58.2 tons). Load cells (0-100 tons, ±0.1 ton accuracy) installed on the bottom of the cars collect actual loads in real time and compare them with calculated values, maintaining a deviation within ±0.5 ton. Generate freight load configuration data including the actual load of each car (57-60 tons), axle weight distribution (20-22 tons / axle), matching coefficient of load and section characteristics (≥0.9), maximum load car number (60 tons for the 10th car), and load adjustment basis (such as the K550-K560 curve section). For example, in the K500-K600 section, the 5th car needs to pass through a 300-meter radius curve, so the load configuration is 58.2 tons, which deviates by 0.1 ton from the calculated value, and the axle weight is 21 tons, which meets the load balancing requirements.

[0166] Step S52: Analyzing the traction power and braking of each train carriage based on the train carriage operation status-train traction force linkage data and the freight load configuration data of each train carriage, and generating traction power-braking data of each train carriage;

[0167] In an embodiment of the present invention, traction power and braking analysis is performed based on the train's operating status and traction linkage data (including adjusted traction and vibration values) of each carriage, combined with freight load configuration data (including the load of each carriage). The traction force value of each carriage in the linkage data (such as 100.8kN for the fifth carriage) and 58.2 tons in the load data are called, and the power demand model is used to calculate the traction power (P=FV, F is the traction force, and V is the scheduled speed of 25m / s). It is concluded that the fifth carriage requires 2520kW. A power sensor (range 0-5000kW, accuracy ±10kW) is installed at the traction motor to collect the actual output power (2515kW), with a deviation of ≤5kW. At the same time, according to the deceleration requirements at the station in the schedule (such as braking starting 1 km before K600 station), the braking distance (800 meters) is calculated in combination with the load data, and the brake cylinder pressure (3MPa corresponds to a deceleration of 0.5m / s) is monitored by a brake pressure sensor (range 0-10MPa, accuracy ±0.05MPa). 2 ) to ensure that the actual braking distance is less than or equal to the calculated value by 10 meters. Generate the traction power of each carriage (2000-3000kW), braking pressure (2-5MPa), power reserve coefficient (actual power / required power ≥ 1.1), braking deceleration (0.3-0.6m / s 2), traction power-brake data of traction-brake switching response time (≤0.5 seconds), for example, the traction power required for the 5th car at K550 is 2520kW, the actual output is 2515kW, the braking pressure is 3.2MPa, and the deceleration is 0.52m / s 2 , which meets the deceleration requirements in the schedule.

[0168] Step S53: Set the train adaptive asynchronous control instructions according to the traction power-braking data of each train car, generate train adaptive asynchronous control instruction data, and execute the train adaptive asynchronous control operation based on the train adaptive asynchronous control instruction data.

[0169] In this embodiment of the present invention, adaptive asynchronous control instructions are set and control operations are executed based on the traction power and brake data (including power and pressure) of each train car. The traction power and brake pressure of each car in the power data are called up, combined with the vibration value (0.6g) and speed difference (0.1km / h) in the linkage data from step S443, and a fuzzy control algorithm is used to generate control instructions: for cars with traction power 5% lower than the required value (for example, the third car with 2400kW requires 2500kW), the motor voltage is commanded to increase by 2% (from 380V to 387.6V); for cars with brake pressure deviation greater than 0.1MPa (for example, the seventh car with 3.1MPa requires 3.0MPa), the brake valve opening is commanded to decrease by 1%. Control commands are sent to each car controller via the train bus (transmission rate 100Mbps, latency ≤10ms). The controller then drives the traction converter and brake valve to perform adjustments. After adjustments, feedback sensors (accuracy ±0.1%) verify that the power of car 3 increases to 2490kW and the pressure of car 7 decreases to 3.02MPa, both meeting the required deviations. Adaptive asynchronous control command data is generated, including the voltage adjustment amount (±1%-3%) for each car, brake valve opening (20%-30%), command execution response time (≤0.2 seconds), post-control parameter deviation (≤2%), and control mode (traction / braking). For example, at K550+500, car 5 experienced a 0.6g vibration, resulting in a 1% reduction in commanded traction force (from 100.8kN to 99.8kN). After execution, the actual traction force was 99.7kN, with a deviation of 0.1kN, achieving precise asynchronous control.

[0170] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0171] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. An adaptive asynchronous control method for railway freight trains based on artificial intelligence, characterized in that: The following steps are involved: Step S1: Collecting train running data through train sensors, performing train operation and track linkage processing based on the train running data, and generating train operation-track linkage data; Step S2: performing real-time train positioning detection based on train running data and train operation-track linkage data to generate real-time train positioning data; Step S3: Acquire train driving planning data; perform train driving schedule design based on the train driving planning data and the train driving real-time positioning data to generate train driving schedule data; Step S4: analyzing the running status of each train carriage and the train traction force linkage based on the real-time train positioning data and the train running-track linkage data, and generating the running status of each train carriage-train traction force linkage data; Step S5: Set the train adaptive asynchronous control instructions based on the operating status of each train car - train traction linkage data and train driving schedule data, generate train adaptive asynchronous control instruction data, and execute the train adaptive asynchronous control operation based on the train adaptive asynchronous control instruction data.

2. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: collecting train running data through train sensors, performing train operation characteristic analysis based on the train running data, and generating train operation characteristic data; Step S12: performing real-time train track detection according to the train operation characteristic data to generate real-time train track data; Step S13: analyzing train track structure change information based on the real-time train running track data to generate train track structure change information data; Step S14: Perform train operation and track linkage processing based on the train operation characteristic data and the train track structure change information data to generate train operation-track linkage data.

3. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing train-tower base station communication and train speed detection according to the train driving data, and generating train-tower base station communication data and train speed data respectively; Step S22: monitoring train running environment changes based on train operation-track linkage data, and generating train running environment change data; Step S23: performing train travel and tunnel travel positioning analysis based on train-tower base station communication data, train speed data, and train travel environment change data to generate train travel-tunnel travel positioning data; Step S24: Perform real-time train positioning detection based on the train-tower base station communication data, train speed data, and train travel-tunnel travel positioning data to generate real-time train positioning data.

4. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 3, characterized in that: Step S23 includes the following steps: Step S231: Analyze the tunnel sequence and time difference of each train carriage according to the train running environment change data to generate tunnel sequence-time difference data of each train carriage; Step S232: Based on the tunnel sequence-time difference data of each train car and the train-tower base station communication data, communication switching and recovery detection of each train car is performed to generate communication switching-recovery data of each train car; Step S233: Calculating the train positioning delay based on the communication switching-restoration data of each train carriage and the train speed data to generate train positioning delay data; Step S234: Perform train travel and tunnel travel positioning analysis based on the train travel positioning delay data and the train speed data to generate train travel-tunnel travel positioning data.

5. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: obtaining train driving planning data, and dividing the train track occupation interval according to the train driving planning data to generate train track occupation interval data; Step S32: Analyze each train intersection track section based on the train track occupation section data and the train driving real-time positioning data to generate each train intersection track section data; Step S33: performing train freight load configuration analysis based on train driving data to generate train freight load configuration data; Step S34: Based on the train freight load configuration data, the intersection track section data of each train, and the train track occupation interval data, the load adaptability matching of each train in the intersection track section is performed to generate the load adaptability data of each train in the intersection track section; Step S35: Based on the train track occupancy section data, the load adaptability data of each train in the intersecting track section, and the intersecting track section data of each train, train schedule design is performed to generate train schedule data.

6. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: performing train track curve and train track gradient detection based on the train real-time positioning data and the train operation-track linkage data, and generating train track curve data and train track gradient data respectively; Step S42: performing a train curve track travel response dynamic analysis based on the train travel track curve data to generate train curve track travel response dynamic data; Step S43: performing a train track gradient response dynamic analysis based on the train track gradient data to generate train track gradient response dynamic data; Step S44: Based on the train curve track travel response power data and the train track gradient response power data, the running status of each train car and the train traction force linkage analysis are performed to generate the running status of each train car-train traction force linkage data.

7. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 6, characterized in that: Step S42 includes the following steps: Step S421: performing train contact axial center of gravity detection on the train operation-track linkage data based on the train track curve data to generate train contact axial center of gravity data; Step S422: performing train track deflection displacement analysis based on the train contact axial center of gravity data to generate train track deflection displacement data; Step S423: performing a train curve track contact area and resistance analysis based on the train track deflection displacement data to generate train curve track contact area-resistance data; Step S424: Perform a train curve track travel response dynamic analysis based on the train curve track contact area-resistance data to generate train curve track travel response dynamic data.

8. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 6, characterized in that: Step S43 includes the following steps: Step S431: Detecting the traction spacing of each carriage on the train operation-track linkage data according to the train track slope data, and generating traction spacing data of each carriage; Step S432: performing push-pull stress analysis on each carriage based on the traction spacing data of each carriage to generate push-pull stress data for each carriage; Step S433: performing axial offset detection on each train based on the push-pull stress data of each carriage and the train operation-track linkage data to generate axial offset data of each train; Step S434: Perform a train track gradient response dynamic analysis based on the axial offset data of each train and the push-pull stress data of each carriage to generate train track gradient response dynamic data.

9. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 6, characterized in that: Step S44 includes the following steps: Step S441: Analyze the longitudinal force coordination and speed difference of each train in the slope section based on the train track gradient response dynamic data, and generate longitudinal force coordination-speed difference data of each train in the slope section; Step S442: Analyze the longitudinal traction deflection force of each train on the curved track based on the train curve track running response dynamic data to generate longitudinal traction deflection force data of each train on the curved track; Step S443: Based on the longitudinal traction deflection force data of each train on the curved track and the longitudinal force coordination-speed difference data of each train in the slope section, the running status of each train car and the train traction force linkage analysis are performed to generate the running status of each train car-train traction force linkage data.

10. The method for adaptive asynchronous control of railway freight trains based on artificial intelligence according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: Calculating the freight load configuration of each train car based on the train schedule data to generate freight load configuration data for each train car; Step S52: Analyzing the traction power and braking of each train carriage based on the train carriage operation status-train traction force linkage data and the freight load configuration data of each train carriage, and generating traction power-braking data of each train carriage; Step S53: Set the train adaptive asynchronous control instructions according to the traction power-braking data of each train car, generate train adaptive asynchronous control instruction data, and execute the train adaptive asynchronous control operation based on the train adaptive asynchronous control instruction data.

Images