Multifunctional track detection system

Abstract

The invention relates to the technical field of rail transit, and provides a multifunctional rail detection system, which comprises a multi-source data acquisition unit, an optical fiber inertial navigation sensor, a laser array module, a dual-mode camera group, a radar speed measurement module and a temperature and humidity sensor, the data acquisition module is used for synchronously acquiring track geometric parameters, three-dimensional surface topography, inertial navigation data, real-time vehicle speed and environmental parameters; the dynamic processing unit is arranged in the compartment and comprises a data fusion controller, a self-adaptive Kalman filtering module, an anomaly diagnosis engine and an edge computing host, and the data fusion controller integrates multi-source data through a multi-thread parallel processing technology. On the basis of self-adaptive Kalman filtering and vehicle speed-parameter dynamic mapping, the track gauge detection error under the variable speed working condition is reduced to + / -0.3 mm from + / -3 mm, and the curvature radius calculation error lt is obtained; and the predictive maintenance model prolongs the overhaul period of the track by 40%, and reduces the operation and maintenance cost by 35%.

Description

Technical Field

[0001] The present invention relates to the field of rail transportation technology, and in particular to a multifunctional rail detection system. Background Art

[0002] The rapid development and intelligentization of rail transit have placed higher demands on track condition monitoring. Traditional detection systems rely on single sensors or manual inspections, making it difficult to achieve all-weather, high-precision, and multi-dimensional monitoring. Data reliability and real-time performance are significantly reduced, especially in complex environments (such as rain, snow, curves, and speed changes).

[0003] Existing technologies include geometric detection systems based on inertial navigation (such as CN113085948B): using an inertial unit (IMU) and laser sensors to measure parameters such as track gauge and height, combined with image-assisted verification, the static accuracy can reach ±1 mm, but serious error accumulation occurs in dynamic environments; and multi-sensor fusion systems (such as CN119428790A): integrating fiber-optic inertial navigation and 3D cameras to achieve signal and image data complementarity, increasing the anomaly detection rate to 95%, but the data processing delay is high (>150 ms), which makes it difficult to meet the real-time needs of high-speed trains.

[0004] The existing technology has a problem when running at variable speeds. The integral errors of geometric parameters accumulate, resulting in a curvature radius calculation deviation of >10%. The image and geometric parameters are processed independently, and the cause of the abnormality cannot be associated (for example, whether the gauge excess is caused by loose fasteners). New sensors need to be rewired, the system upgrade cost is high, and the transmission delay of large data volumes is >200ms, which affects emergency response. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides a multifunctional track detection system, which solves the problems of accumulated integral errors of geometric parameters during variable speed operation, resulting in a curvature radius calculation deviation of >10%, independent processing of images and geometric parameters, and inability to associate the cause of the abnormality (such as whether the gauge excess is caused by loose fasteners). New sensors need to be rewired, the system upgrade cost is high, and the transmission delay of large data volumes is >200ms, which affects emergency response.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a multifunctional track detection system, comprising: A multi-source data acquisition unit installed on the bottom of the train, consisting of a fiber-optic inertial navigation sensor, a laser array module, a dual-mode camera set, a radar speed measurement module, and a temperature and humidity sensor, is used to simultaneously collect track geometry parameters, three-dimensional surface topography, inertial navigation data, real-time vehicle speed, and environmental parameters; The dynamic processing unit set in the car includes a data fusion controller, an adaptive Kalman filter module, an abnormality diagnosis engine and an edge computing host. The data fusion controller integrates multi-source data through multi-threaded parallel processing technology to generate a track state matrix M = [P 几何 , I 图像 , A 惯性 ], where P 几何 Including gauge, height, and track direction parameters, I 图像 is the track surface defect eigenvector, A 惯性 is the acceleration and attitude angle data; The power management unit adopts a multi-level isolated power supply architecture to convert the train's DC110V input into 24V, 5V, and 3.3V outputs. It integrates anti-reverse connection circuits and surge protectors, supports independent power-off and restart of each module, and the restart delay time is t d =0.1±0.02s; The network communication unit includes a dual-redundant Gigabit Ethernet switch and an RS-485 / 232 protocol conversion module, enabling full-duplex data transmission from the bottom of the vehicle to the interior of the vehicle; The remote interaction unit, equipped with an AR augmented reality interface and AI decision-making module, displays track health maps in real time and generates maintenance work orders.

[0007] Preferably, the dual-mode camera group includes: The 2D linear array camera on the rail surface captures rail surface images at a sampling interval of 1.6 mm and combines it with a convolutional neural network to identify cracks, scratches, and corrugation defects. The identification threshold δ 缺陷 ≥0.5mm; The 3D structured light camera for roadbed fasteners uses 850nm laser-assisted lighting to scan the 3D point cloud data of roadbed fasteners and calculate the fastener offset using a point cloud registration algorithm. , accuracy reaches ±0.05mm.

[0008] Preferably, the laser array module comprises: A transverse laser transmitter projects laser lines onto the track section at 0.25m intervals to generate a dynamic profile map; Longitudinal laser radar, measuring the longitudinal slope α of the track, and calculating the track curvature radius in combination with inertial data ,in is the vehicle speed, is the tilt sensor data, is the acceleration due to gravity.

[0009] Preferably, the adaptive Kalman filter module performs the following steps: Step 1: Input multi-source data 几何 , I 图像 , A 惯性}; Step 2: Construct the state equation X k =F˙X k−1 +B˙u k +w k , where F is the state transfer matrix, u k is the control variable, w k is the process noise; Step 3: Through the observation matrix Z k =H˙X k +v k Correct the predicted value and output the filtered geometric parameter P 滤波 =P 原始 ˙(1−k v ˙Δv), where k v is the vehicle speed correction coefficient, and Δv is the vehicle speed change.

[0010] Preferably, the abnormality diagnosis engine includes: The track gauge overrun detection module triggers image correlation analysis when the track gauge L>1435±2mm, using a 3D camera to verify whether it is a turnout area or a loose fastener. Wear warning module, if the track profile wear δ=|L 实测 −L 标准 ∣≥1.5 mm, generate maintenance coordinates (X, Y, Z) and mark the wear type.

[0011] Preferably, the network communication unit adopts: Time-sensitive network protocol, priority allocation image data bandwidth ≥ 800Mbps, geometry parameter data bandwidth ≥ 200Mbps; Fiber optic redundant ring network, switching time ≤ 10ms in case of single point failure, ensuring data transmission continuity.

[0012] Preferably, the power management unit integrates: Energy consumption controller, dynamically adjusts the sensor sampling frequency f according to the vehicle speed v s , when v < 50 km / h f s =100Hz, when v≥200 km / h, f s =500 Hz; Supercapacitor energy storage module to maintain system operation when train power supply is interrupted 后备 ≥5 mint.

[0013] Preferably, the remote interaction unit supports: Multi-dimensional data visualization, superimposing track geometry deviations, surface defects, and environmental parameters onto a 3D digital twin model; Predictive maintenance algorithm, based on LSTM neural network to predict the remaining life of the track ,in is the wear weight coefficient.

[0014] Preferably, the edge computing host is connected to the cloud platform via 5G communication to achieve: Distributed data storage, track health data is encrypted and uploaded in shards, and supports blockchain evidence storage; Multi-train collaborative analysis and optimization of the global track status prediction model through federated learning.

[0015] Preferably, adaptive environmental compensation is also included: When the temperature and humidity sensor detects a rainy or foggy environment, it activates the laser scattering compensation algorithm to correct image noise; In low temperature environments, the camera lens is automatically heated to prevent frost.

[0016] Working principle: Using a precise clock protocol, the clock deviation of all sensors is ≤1μs, ensuring the consistency of data timestamps; the laser and camera are jointly calibrated using a calibration board, and the spatial coordinate error is ≤0.1 mm.

[0017] Dynamic data fusion and compensation: Establish vehicle speed v and track gauge correction coefficient k v Mapping table; Dynamically adjust track gauge L 校正 =L 原始 ˙k v , eliminating the measurement deviation caused by centrifugal force.

[0018] Process noise covariance Q=diag(0.1 2 ,0.1 2 ), observation noise covariance R=diag(0.05 2 ); Output filtered geometric parameters P 滤波 , the standard deviation is reduced to 0.3 mm.

[0019] Response to extreme environments: Rain and fog mode: The laser power is automatically increased by 20% to compensate for scattering loss; the image algorithm uses the defogging model I 清晰 =(I 原始 −A)˙t −1 +A, where A is the atmospheric light intensity and t is the transmittance.

[0020] Low temperature mode: The supercapacitor powers the heating film to maintain the lens temperature; the data storage module switches to low-temperature resistant flash memory (operating range -40°C to 85°C).

[0021] The present invention provides a multifunctional track detection system. It has the following beneficial effects: 1. Based on adaptive Kalman filtering and dynamic vehicle speed-parameter mapping, the present invention reduces the track gauge detection error under variable speed conditions from ±3mm to ±0.3mm, and the curvature radius calculation error is less than 2%. The predictive maintenance model extends the track overhaul cycle by 40% and reduces operation and maintenance costs by 35%.

[0022] 2. This invention uses laser scattering compensation and lens heating technology to increase the image signal-to-noise ratio (SNR) to 25dB in rainy and foggy environments, and the low-temperature data availability rate is >95%. By combining the TSN protocol with edge computing, the data processing delay is compressed to 30ms, supporting real-time monitoring of trains with speeds ≥400km / h. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A system diagram of the present invention; Figure 2 System diagram of the power management unit of the present invention; Figure 3 A system diagram of the remote interaction unit of the present invention; Figure 4 A system diagram of the dynamic processing unit of the present invention; Figure 5 This is a system diagram of multiple data acquisition units in the present invention. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] As one aspect of the present invention, please refer to the attached Figure 1 , an embodiment of the present invention provides a multifunctional track detection system, comprising: Please see the attached Figure 5 The multi-source data acquisition unit installed on the bottom of the train includes a fiber optic inertial navigation sensor, a laser array module, a dual-mode camera group, a radar speed measurement module, and a temperature and humidity sensor. It is used to synchronously collect track geometry parameters, three-dimensional surface topography, inertial navigation data, real-time vehicle speed, and environmental parameters. The dual-mode camera set includes: The rail surface 2D linear array camera captures rail surface images at a sampling interval of 1.6mm and combines it with a convolutional neural network to identify cracks, scratches, and corrugation defects, with a recognition threshold of δ defects ≥ 0.5mm. The 3D structured light camera for roadbed fasteners uses 850nm laser-assisted lighting to scan the 3D point cloud data of roadbed fasteners and calculate the fastener offset using a point cloud registration algorithm. , accuracy reaches ±0.05mm; The laser array module includes: A transverse laser transmitter projects laser lines onto the track section at 0.25m intervals to generate a dynamic profile map; Longitudinal laser radar, measuring the longitudinal slope α of the track, and calculating the track curvature radius in combination with inertial data ,in is the vehicle speed, is the tilt sensor data, is the acceleration due to gravity; Please see the attached Figure 4 The dynamic processing unit installed in the carriage includes a data fusion controller, an adaptive Kalman filter module, an abnormality diagnosis engine and an edge computing host. The data fusion controller integrates multi-source data through multi-threaded parallel processing technology to generate a track state matrix M = [P geometry, I image, A inertia], where P geometry includes gauge, height and track direction parameters, I image is the track surface defect feature vector, and A inertia is the acceleration and attitude angle data; Among them, the anomaly diagnosis engine includes: The track gauge overrun detection module triggers image correlation analysis when the track gauge L>1435±2mm, using a 3D camera to verify whether it is a turnout area or a loose fastener. Wear warning module: if the rail profile wear δ = |L measured − L standard | ≥ 1.5 mm, it generates maintenance coordinates (X, Y, Z) and marks the wear type; The network communication unit adopts: Time-sensitive network protocol, priority allocation image data bandwidth ≥ 800Mbps, geometry parameter data bandwidth ≥ 200Mbps; Fiber optic redundant ring network, switching time ≤ 10ms in case of single point failure, ensuring data transmission continuity; The edge computing host and the cloud platform are connected via 5G communication to achieve: Distributed data storage, track health data is encrypted and uploaded in shards, and supports blockchain evidence storage; Multi-train collaborative analysis, optimizing the global track status prediction model through federated learning; Please see the attached Figure 2, adopts a power management unit with a multi-level isolated power supply architecture to convert the train DC110V input into 24V, 5V, and 3.3V outputs, integrates anti-reverse connection circuit and surge protector, supports independent power off and restart of each module, and the restart delay time is t d =0.1±0.02s, where the power management unit integrates: Energy consumption controller, dynamically adjusts the sensor sampling frequency f according to the vehicle speed v s , when v < 50 km / h f s =100Hz, when v≥200 km / h, f s =500 Hz; Supercapacitor energy storage module, maintaining system operation with a backup time of ≥5 mint when the train power supply is interrupted; The network communication unit includes a dual-redundant Gigabit Ethernet switch and an RS-485 / 232 protocol conversion module, enabling full-duplex data transmission from the bottom of the vehicle to the interior of the vehicle; Please see the attached Figure 3 The remote interaction unit, equipped with an AR augmented reality interface and AI decision-making module, displays track health maps in real time and generates maintenance work orders. The remote interaction unit supports: Multi-dimensional data visualization, superimposing track geometry deviations, surface defects, and environmental parameters onto a 3D digital twin model; Predictive maintenance algorithm, based on LSTM neural network to predict the remaining life of the track ,in is the wear weight coefficient; Adaptive environmental compensation: When the temperature and humidity sensor detects a rainy or foggy environment, it activates the laser scattering compensation algorithm to correct image noise; In low temperature environments, the camera lens is automatically heated to prevent frost.

[0026] As another aspect of the present invention, in the multifunctional track detection system provided above, the adaptive Kalman filter module performs the following steps: Step 1: Input multi-source data {P geometry, I image, A inertia}; Step 2: Construct the state equation Xk=F˙Xk−1+B˙uk+wk, where F is the state transfer matrix, uk is the control variable, and wk is the process noise; Step 3: Correct the predicted value using the observation matrix Zk = H˙Xk + vk, and output the filtered geometric parameters Pfiltered = Poriginal˙(1−kv˙Δv), where kv is the vehicle speed correction coefficient and Δv is the vehicle speed change.

[0027] The following is an introduction with reference to specific embodiments: Example Train installation: The system is deployed on a certain type of high-speed EMU (operating speed 350km / h). A multi-source data acquisition unit is installed on the bottom of the train, and a dynamic processing unit and a remote interactive terminal are set up in the carriage.

[0028] Sensor layout: Fiber optic inertial navigation sensor: installed on the center line of the vehicle bottom, with a sampling frequency of 500Hz, measuring the three-dimensional acceleration of the train (a x ,a y ,a z ) and angular velocity; Laser array module: The horizontal laser scans the track section at 0.25m intervals, and the longitudinal laser radar measures the track slope (resolution 0.01°); Dual-mode camera system: a 2D line scan camera (resolution 10μm / pixel) and a 3D structured light camera (point cloud density 1000 points / cm²) are aimed at the rail surface and the trackbed fasteners respectively; Temperature and humidity sensor: integrated into the vehicle bottom protective cover, monitors the ambient humidity and temperature in real time.

[0029] Data collection: Geometric parameters: The laser sensor generates a track profile in real time, which is compared with the standard profile to calculate the wear loss δ = |L measured − L standard |; Image data: A 2D camera captures rail surface images and uses a CNN algorithm to identify cracks (with a threshold of 0.5 mm). A 3D camera scans the fastener point cloud and calculates the offset Δd. Inertial data: Fiber optic inertial navigation outputs acceleration and attitude angle, combined with vehicle speed v to calculate track curvature radius .

[0030] Data transmission: The vehicle bottom data is transmitted to the dynamic processing unit in the vehicle cabin via dual redundant Gigabit Ethernet (TSN protocol). Priority allocation ensures that the image data bandwidth is ≥800Mbps and the transmission delay is ≤20ms. The RS-485 protocol conversion module uploads the temperature and humidity data to the edge computing host.

[0031] Data processing: Adaptive Kalman filter: Input multi-source data to construct state equation X k =F˙X k−1 +B˙u k +w k , through the observation matrix Z k =H˙X k +v k Correct geometric parameters and dynamically compensate for vehicle speed variation errors; Abnormal diagnosis: If the track gauge L is greater than 1435±2 mm, the system correlates the 3D fastener image to determine whether it is loose (offset Δd>1 mm) or in the turnout area.

[0032] Rainy and foggy environment: When the humidity is ≥90%, the laser scattering compensation algorithm is activated and the filter function I 校正 =I 原始 ˙e −k˙湿度 Correct image noise and increase the signal-to-noise ratio (SNR) to 25dB; Low temperature environment: When the temperature sensor detects T<−20°C, the camera lens heating film is automatically activated with a power of P=5 W˙(T 目标 −T 实际 ) Ensure that the lens temperature is ≥0℃.

[0033] Remote interactive terminal: The AR interface overlays a real-time track health map, annotates the coordinates of abnormal points, and generates a repair work order. Fasteners are loose and torque calibration to 120 N·m is recommended. Cloud collaboration: Data is encrypted and uploaded to the blockchain platform in shards, supporting multi-train federated learning to optimize the global model.

[0034] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A multifunctional track detection system, characterized in that: include: A multi-source data acquisition unit installed on the bottom of the train, consisting of a fiber-optic inertial navigation sensor, a laser array module, a dual-mode camera set, a radar speed measurement module, and a temperature and humidity sensor, is used to simultaneously collect track geometry parameters, three-dimensional surface topography, inertial navigation data, real-time vehicle speed, and environmental parameters; The dynamic processing unit set in the car includes a data fusion controller, an adaptive Kalman filter module, an abnormality diagnosis engine and an edge computing host. The data fusion controller integrates multi-source data through multi-threaded parallel processing technology to generate a track state matrix M = [P 几何 , I 图像 , A 惯性 ], where P 几何 Including gauge, height, and track direction parameters, I 图像 is the track surface defect feature vector, A 惯性 is the acceleration and attitude angle data; The power management unit adopts a multi-level isolated power supply architecture to convert the train's DC110V input into 24V, 5V, and 3.3V outputs. It integrates anti-reverse connection circuits and surge protectors, supports independent power-off and restart of each module, and the restart delay time is t d =0.1±0.02s; The network communication unit includes a dual-redundant Gigabit Ethernet switch and an RS-485 / 232 protocol conversion module, enabling full-duplex data transmission from the bottom of the vehicle to the interior of the vehicle; The remote interaction unit, equipped with an AR augmented reality interface and AI decision-making module, displays track health maps in real time and generates maintenance work orders.

2. A multifunctional track detection system according to claim 1, characterized in that: The dual-mode camera set includes: The 2D linear array camera on the rail surface captures rail surface images at a sampling interval of 1.6 mm and combines it with a convolutional neural network to identify cracks, scratches, and corrugation defects. The identification threshold δ 缺陷 ≥0.5mm; The 3D structured light camera for roadbed fasteners uses 850nm laser-assisted lighting to scan the 3D point cloud data of roadbed fasteners and calculate the fastener offset using a point cloud registration algorithm. , accuracy reaches ±0.05mm.

3. A multifunctional track detection system according to claim 1, characterized in that: The laser array module comprises: A transverse laser transmitter projects laser lines onto the track section at 0.25m intervals to generate a dynamic profile map; Longitudinal laser radar, measuring the longitudinal slope α of the track, and calculating the track curvature radius in combination with inertial data ,in is the vehicle speed, is the tilt sensor data, is the acceleration due to gravity.

4. A multifunctional track detection system according to claim 1, characterized in that: The adaptive Kalman filter module performs the following steps: Step 1: Input multi-source data 几何 , I 图像 , A 惯性 }; Step 2: Construct the state equation X k =F˙X k−1 +B˙u k +w k , where F is the state transfer matrix, u k is the control variable, w k is the process noise; Step 3: Through the observation matrix Z k =H˙X k +v k Correct the predicted value and output the filtered geometric parameter P 滤波 =P 原始 ˙(1−k v ˙Δv), where k v is the vehicle speed correction coefficient, and Δv is the vehicle speed change.

5. The multifunctional track detection system according to claim 1, characterized in that: The abnormality diagnosis engine includes: The track gauge overrun detection module triggers image correlation analysis when the track gauge L>1435±2mm, using a 3D camera to verify whether it is a turnout area or a loose fastener. Wear warning module, if the track profile wear δ=|L 实测 −L 标准 ∣≥1.5 mm, generate maintenance coordinates (X, Y, Z) and mark the wear type.

6. A multifunctional track detection system according to claim 1, characterized in that: The network communication unit adopts: Time-sensitive network protocol, priority allocation image data bandwidth ≥ 800Mbps, geometry parameter data bandwidth ≥ 200Mbps; Fiber optic redundant ring network, switching time ≤ 10ms in case of single point failure, ensuring data transmission continuity.

7. The multifunctional track detection system according to claim 1, characterized in that: The power management unit integrates: Energy consumption controller, dynamically adjusts the sensor sampling frequency f according to the vehicle speed v s , when v < 50 km / h f s =100 Hz, when v≥200 km / h, f s =500 Hz; Supercapacitor energy storage module to maintain system operation when train power supply is interrupted 后备 ≥5 mint.

8. The multifunctional track detection system according to claim 1, characterized in that: The remote interaction unit supports: Multi-dimensional data visualization, superimposing track geometry deviations, surface defects, and environmental parameters onto a 3D digital twin model; Predictive maintenance algorithm, based on LSTM neural network to predict the remaining life of the track ,in is the wear weight coefficient.

9. The multifunctional track detection system according to claim 1, characterized in that: The edge computing host and the cloud platform are connected via 5G communication to achieve: Distributed data storage, track health data is encrypted and uploaded in shards, and supports blockchain evidence storage; Multi-train collaborative analysis and optimization of the global track status prediction model through federated learning.

10. The multifunctional track detection system according to claim 1, characterized in that: Also includes adaptive environmental compensation: When the temperature and humidity sensor detects a rainy or foggy environment, it activates the laser scattering compensation algorithm to correct image noise; In low temperature environments, the camera lens is automatically heated to prevent frost.

Images