A multi-dimensional pose decoupling twds online detection dynamic compensation method and system
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHINA ACADEMY OF RAILWAY SCI CORP LTD
- Filing Date
- 2026-05-09
- Publication Date
- 2026-06-16
Smart Images
Figure CN122217166A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of rail transit inspection technology, and more specifically, to a TWDS online detection dynamic compensation method and system with multidimensional pose decoupling. Background Technology
[0002] The wheelset dimensional inspection system for freight cars is a key piece of equipment for online non-contact measurement of wheelset geometric parameters in rail transit vehicles, and its measurement accuracy directly affects the reliability of driving safety assessment. Existing TWDS (Transportation Wheelset Data System) solutions, such as CN106091951A, employ a dual-laser displacement sensor arrangement with three magnets to calculate the flange height and flange thickness by acquiring the two-dimensional profile curve of the wheel; CN119756182A further proposes a modular data acquisition system to achieve automated acquisition and processing of wheelset geometric data.
[0003] However, the common limitation of the above schemes is that they all assume that the wheel moves in an ideal straight line within the measurement area, without considering the serpentine oscillation induced by the interaction between the wheel and rail in actual operation—that is, the wheel simultaneously undergoes lateral displacement, yaw vibration around the vertical axis, and roll vibration around the horizontal axis, forming a complex motion with multiple degrees of freedom coupling.
[0004] The fundamental impact of serpentine oscillation on the accuracy of TWDS measurements is as follows: First, the continuous time-varying spatial pose of the wheel causes nonlinear errors in the original contour point cloud; second, the dynamic change of the laser projection angle disrupts the geometric projection relationship, producing projection distortions that are physically unobservable and cannot be corrected by conventional fitting algorithms.
[0005] More importantly, existing technologies all employ a holistic compensation strategy, treating composite motion as a single error source and uniformly correcting it through a single correction parameter. The inherent flaw of this strategy lies in its inability to distinguish the independent contributions of each motion dimension, such as lateral movement, yaw, and roll, to the measurement error. Errors in each dimension are severely coupled, the compensation accuracy is limited by the most unfavorable error component, and the accuracy of each dimension cannot be independently verified. Furthermore, it is impossible to quickly locate and isolate anomalies.
[0006] Therefore, under conditions such as high speed and heavy load, which easily induce swaying, the measurement error of wheel flange parameters increases significantly with existing technology, making it difficult to meet the increasingly stringent requirements for wheelset safety monitoring. Summary of the Invention
[0007] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a multidimensional pose decoupling TWDS online detection dynamic compensation method and system to solve the problems mentioned in the background art.
[0008] To achieve the above objectives, the present invention provides the following technical solution: a TWDS online detection dynamic compensation method with multidimensional pose decoupling, comprising the following steps:
[0009] S1. Deploy multi-dimensional pose sensing units along the track measurement area. The multi-dimensional pose sensing unit includes at least three sets of laser displacement sensors, two sets of inertial measurement units and four induction magnets, used to collect the original two-dimensional contour point cloud of the wheel, the three-axis acceleration and three-axis angular velocity of the sensor bracket, and the wheel through the trigger time stamp, and to perform unified spatial coordinate system calibration on all sensors.
[0010] S2. When the train passes through the measurement area, an accurate timestamp is generated using an induction magnet to synchronize and align the collected raw point cloud frame with the inertial data in time. The instantaneous pose vector of the support is obtained by integrating the inertial data to eliminate the physical vibration deviation of the measurement platform.
[0011] S3. Based on redundant observations from multiple laser sensors and magnetic positioning signals, the composite motion of the wheel relative to the track centerline is decoupled into independent pose components of lateral displacement, yaw angle, and roll angle; a 6-DOF dynamic transformation matrix is constructed for each sampling moment through error components of each dimension.
[0012] S4. The original point cloud is inversely transformed using the dynamic transformation matrix to correct the projection geometric distortion caused by the serpentine swing; the corrected cross-sectional data of multiple frames are spatially fused to establish a high-precision three-dimensional mathematical model of the wheel and its flange area, and key geometric indicators are automatically extracted from the three-dimensional model to detect wheel defects.
[0013] S5. Compare the corrected data measured by multiple sets of sensors to determine the convergence and consistency of the measurement results. When the calculation results exceed the preset industry safety tolerance, a real-time alarm is triggered, and digital maintenance suggestions and evaluation reports are output.
[0014] Preferably, as a preferred embodiment of the TWDS online detection dynamic compensation method for multidimensional pose decoupling described in this invention, it includes, in step S1, three sets of laser displacement sensors are sequentially installed on the outer and inner sides of the track along the rail direction to acquire the original contour point cloud data of the wheel tread and wheel flange areas; two sets of inertial measurement units are installed on the rigid support where the laser displacement sensors are located to measure the three-axis acceleration and three-axis angular velocity of the support in real time when the train passes; and four induction magnets are sequentially installed on the inner side of the track along the rail direction to detect the rail direction position of the wheel center and generate a trigger time stamp signal.
[0015] All sensors are calibrated using a unified spatial coordinate system. The origin is the track centerline, the positive X-axis is along the track direction, the positive Y-axis is perpendicular to the track surface to the left, and the positive Z-axis is perpendicular to the track plane upwards. A world coordinate system is then established with the track centerline as the reference. The local point cloud coordinates of the three sets of laser displacement sensors, the inertial measurement vectors of the two sets of inertial measurement units, and the track orientation information of the magnet are uniformly converted to [the correct coordinates] using a calibration matrix. In the coordinate system, the static system error introduced by the differences in installation position and angle of each sensor is eliminated.
[0016] Preferably, as a preferred embodiment of the TWDS online detection dynamic compensation method for multidimensional pose decoupling described in this invention, it includes, in step S2, when the train enters the measurement area, four induction magnets are triggered sequentially to generate a precise timestamp sequence of the wheels reaching each preset position. The generated precise timestamp sequence is denoted as... ,in, These represent the moments when the wheelset is detected entering the vehicle and when the protective gate is triggered to open, respectively. This indicates the measurement reference time, corresponding to the moment when the wheel center projection reaches directly below the measurement module. This indicates the moment when the wheelset leaves the measurement area and triggers the protection system to reset;
[0017] The rising edge of the trigger of the third induction magnet To serve as a synchronization reference, the raw point cloud frames acquired by the laser displacement sensor and the inertial data acquired by the inertial measurement unit are synchronized at the microsecond level.
[0018] The triaxial acceleration and triaxial angular velocity output from the inertial measurement unit are extracted, and the instantaneous spatial displacement vector and deflection vector of the laser displacement sensor bracket when the train passes are obtained through integration.
[0019] By using instantaneous spatial displacement vector and deflection vector, a support vibration compensation matrix is established, and the original two-dimensional contour point cloud is corrected by inverse coordinate transformation. By compensating for the instantaneous displacement and tilt of the support, the support excitation error induced by the high-speed operation of the train is eliminated, and the original wheel point cloud relative to the static track reference frame is obtained.
[0020] Preferably, as a preferred embodiment of the TWDS online detection dynamic compensation method for multidimensional pose decoupling described in this invention, it includes, in step S3, decoupling the composite motion of the wheel relative to the track centerline into independent pose components of lateral displacement, yaw angle, and roll angle based on redundant observations from multiple sets of laser sensors and magnetic positioning signals; constructing a 6-DOF dynamic transformation matrix for each sampling moment through the error components of each dimension; specifically including the following:
[0021] By utilizing redundant observation sections formed by three sets of laser displacement sensors along the track direction, combined with the real-time wheel center position determined by induction magnets, a geometric analytical algorithm decouples the complex serpentine oscillation of the wheel relative to the track centerline into independent pose components, including lateral displacement, yaw angle, and roll angle. The lateral displacement is extracted based on the change in lateral distance between the wheel flange and the tread measured by the inner and outer sensors. The yaw angle is calculated by using the observation time difference and distance deviation of the same wheelset side by three sets of scanners distributed along the track direction to determine the wheel's yaw angle around the vertical axis. The roll angle is extracted by analyzing the vertical height difference distribution of tread slope points to determine the wheel's lateral tilt angle relative to the track plane.
[0022] Independent error components in each dimension are obtained by decoupling the lateral swing and head-shaking angle coupling and the vertical jump and side roll motion decoupling, and a 6-DOF dynamic transformation matrix of the scanned point cloud is constructed at each moment.
[0023] The decoupling of lateral sway and head-tilt angle coupling is achieved by using three sets of laser displacement sensors with fixed installation spacing in the rail direction, and comparing the lateral displacement values of the tread obtained by different sensors at the same time. Obtain the pure lateral displacement vector. and yaw angle in radians ,in, This is a correction term for the geometric projection deviation caused by the yaw angle. These represent the lateral displacement observations of the wheel tread obtained by different sensors at the same time, where L is the installation spacing of the sensors along the rail direction;
[0024] The decoupling of vertical bounce and lateral roll motion is achieved by using a point cloud of lateral feature points on the tread acquired by sensors to fit the lateral slope of the tread. And combined with the standard tread slope To distinguish between vertical displacement and tilt angle, the specific formula is as follows: , ,in, R is the observed vertical average height, and R is the nominal radius of the wheel. Let θ be the tilt angle and z be the vertical displacement.
[0025] For the scanned point cloud acquired at each moment, the decoupled independent components are integrated. and limited The pitch angle is calculated by integrating the angular velocity around the Y-axis in real time using an inertial measurement unit mounted on a support bracket beside the track, thus constructing a 6-DOF dynamic transformation matrix for each moment. The dynamic transformation matrix The specific expression is as follows .
[0026] Preferably, as a preferred embodiment of the TWDS online detection dynamic compensation method for multidimensional pose decoupling described in this invention, it includes, in step S4, performing an inverse transformation on the original point cloud using the dynamic transformation matrix to correct the projection geometric distortion caused by serpentine swaying; specifically including:
[0027] Using the 6-DOF dynamic transformation matrix constructed in step S3, an inverse spatial transformation is performed on each frame of the original point cloud to obtain the corrected standard meridional point cloud. ,in, The original point cloud is in homogeneous coordinate form. The coordinates of the standard meridian point cloud are the corrected coordinates. It is the inverse matrix of the dynamic transformation matrix; by calculating and compensating for the yaw, roll and lateral displacement deviations of the wheel at the moment of measurement, the "pseudo-distorted profile" is restored to the meridional point cloud that reflects the true cross section of the wheel;
[0028] The step of spatially fusing multi-frame corrected cross-sectional data to establish a high-precision three-dimensional mathematical model of the wheel and its flange region, and automatically extracting key geometric indicators and detecting wheel defects from the three-dimensional model, further includes:
[0029] Multiple frames of corrected two-dimensional cross-sectional data continuously collected during the train's passage are precisely registered and fused according to the spatial position sequence determined by the magnet in step S2. The spatial position sequence of each frame is determined based on the magnet positioning signal, and the spatial position of the i-th frame is represented as... ,in, The starting position is given by v, where v is the wheel speed, calculated by dividing the distance between adjacent magnets by the time difference. The acquisition time of the i-th frame;
[0030] The multi-frame corrected 2D cross-sectional point clouds are stitched together according to their corresponding spatial positions to obtain a fused point cloud set. The fused point cloud is then reconstructed using a moving least squares surface fitting algorithm to generate a continuous and smooth 3D tread surface model. ;
[0031] Key geometric parameters, including rim thickness, rim height, rim slope, and tread wear, are automatically identified and extracted from the 3D model. Wheel defects, such as scratches, pore depth, and pore area, are identified and detected using the variation characteristics of the surface normal vector and curvature of the 3D model. Curvature abrupt change detection technology using 3D point clouds is employed to calculate the average curvature of each point on the model surface. ,in, These are the maximum and minimum principal curvatures of the surface at point p, respectively;
[0032] And calculate the normal vector of each point p on the surface of the 3D model. The mean curvature Reference curvature of the standard tread area When comparing, and When, point p is marked as a candidate defect point, where, The reference curvature for the healthy tread surface area. Let be the curvature abrupt change threshold, and q be a neighborhood point of p. The threshold for the change in the normal vector;
[0033] Cluster analysis is performed on the marked defect candidate points to segment independent defect regions. Based on the morphological characteristics of the defect regions, scratches and pores are distinguished: those with open boundaries and an area greater than a preset threshold are judged as scratches, and those with closed hole boundaries are judged as pores.
[0034] For the scratched area, calculate the projected area A and the maximum depth. ,when The system immediately detected that the wheel had skid damage and issued a warning. The standard tread profile surface is used; the method for calculating the scratch area A is as follows: the scratch area point set is triangulated, and the total area is obtained by summing the areas of each triangle;
[0035] For the pore region, identify the defect region with closed pore boundary, calculate the normal distance from the deepest point in the pore to the standard tread profile as the pore depth, and triangulate the pore opening region to obtain the pore area.
[0036] The maximum depth The calculation method is as follows: calculate the normal distance from each point in the scratch area to the fitted surface of the standard tread profile, and take the maximum value.
[0037] Preferably, as a preferred embodiment of the TWDS online detection dynamic compensation method for multi-dimensional pose decoupling described in this invention, it includes, in step S5, when the calculation result exceeds the preset industry safety tolerance, automatically issuing a real-time alarm at the red warning level, comparing the measured three-dimensional profile with the standard template profile, automatically calculating the minimum wheel repair depth value of the wheelset under the premise of meeting safety standards, and outputting digital maintenance suggestions and vehicle operating status assessment reports, specifically including the following:
[0038] Mutual inductance verification is performed on the geometric parameters of the three sets of laser displacement sensors after independent calculation and correction to determine the convergence of the measurement results. When the data deviation between each group is within the preset threshold, the weighted average value is taken as the final measured value to ensure the stability and reliability of the output data.
[0039] The extracted key geometric indicators are compared with the preset industry safety tolerances in real time. The safety tolerances include the upper limit of the flange height, the lower limit of the flange thickness, and the upper limit of the tread wear. When the calculated result of any indicator exceeds the preset industry safety upper limit, the system automatically determines it to be a red warning level and triggers a real-time alarm signal to notify the dispatching department to carry out vehicle impoundment inspection.
[0040] The measured 3D profile reconstructed by step S4 Contour of standard template Spatial registration and interpolation are performed, and the globally optimal cutting path is found through an iterative algorithm. The minimum overhaul depth of the wheel is automatically calculated. ,in, The target outline after rotation modification;
[0041] It automatically summarizes measurement results, including: rim height, rim thickness, tread wear, scratch detection results, and minimum repair depth, and outputs a digital assessment report including minimum repair recommendations, expected post-repair life, and overall vehicle operating safety status to guide repair workshops in making precise repairs.
[0042] This application also provides a TWDS online detection dynamic compensation system with multi-dimensional pose decoupling, specifically including a multi-dimensional pose perception unit, a time synchronization and compensation unit, a multi-dimensional pose decoupling unit, a three-dimensional reconstruction unit, a defect detection unit, and an early warning and decision output unit;
[0043] The multidimensional pose sensing unit consists of at least three sets of laser displacement sensors, two sets of inertial measurement units, and four inductive magnets. It is deployed along the track measurement area to collect the original two-dimensional contour point cloud of the wheel, the three-axis acceleration and three-axis angular velocity of the sensor bracket, and the trigger timestamp of the wheel. All sensors in the sensing unit are calibrated using a unified spatial coordinate system.
[0044] The time synchronization and compensation unit is electrically connected to the multi-dimensional pose sensing unit. It is used to generate a precise timestamp using an inductive magnet, synchronize and align the collected original point cloud frame with the inertial data, and obtain the instantaneous pose vector of the support through inertial data integration to eliminate the physical vibration deviation of the measurement platform.
[0045] The multi-dimensional pose decoupling unit is connected to the time synchronization and compensation unit. It is used to decouple the composite motion of the wheel relative to the center line of the track into independent pose components of lateral displacement, yaw angle and roll angle based on redundant observations from multiple sets of laser sensors and magnetic positioning signals, and to construct a 6-DOF dynamic transformation matrix for each sampling time.
[0046] The three-dimensional reconstruction unit is connected to the multi-dimensional pose decoupling unit and is used to perform inverse transformation processing on the original point cloud using the dynamic transformation matrix, correct the projection geometric distortion caused by the serpentine swing, and spatially fuse the corrected cross-sectional data of multiple frames to establish a high-precision three-dimensional mathematical model of the wheel and its flange area.
[0047] The defect detection unit is connected to the three-dimensional reconstruction unit and is used to automatically extract key geometric indicators from the three-dimensional model and detect wheel defects.
[0048] The early warning and decision output unit is connected to the defect detection unit and is used to compare the corrected data measured by multiple sets of sensors to determine the convergence and consistency of the measurement results. When the calculation results exceed the preset industry safety tolerance, a real-time alarm is triggered, and digital maintenance suggestions and evaluation reports are output.
[0049] On the other hand, the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements a functional module of a multidimensional pose decoupling TWDS online detection dynamic compensation method as described above.
[0050] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements a multidimensional pose decoupling TWDS online detection dynamic compensation method as described above.
[0051] The technical effects and advantages provided by the present invention in the above technical solution are as follows:
[0052] This invention significantly improves the accuracy and intelligence of online detection in rail transit through multi-source sensor fusion and dynamic compensation algorithms. Addressing the projected geometric distortion caused by swaying, it constructs a 6-DOF dynamic transformation matrix to achieve physical-level source tracing and inverse transformation restoration of measurement distortion, solving the problem of nonlinear measurement error in wheel flange thickness parameters. An inertial measurement unit is introduced to offset support vibration deviation in real time. Combined with precise magnetic alignment technology, it effectively separates track elastic sinking from equipment vibration interference, ensuring data stability and high reliability under complex dynamic conditions. This invention can effectively eliminate measurement platform vibration and swaying interference, significantly improving online detection accuracy. Attached Figure Description
[0053] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0054] Figure 1 This is a flowchart of the method of the present invention.
[0055] Figure 2 This is a schematic diagram of the layout of the multi-dimensional pose sensing unit.
[0056] Figure 3 This is a comparison image before and after point cloud correction.
[0057] Figure 4 A schematic diagram of the mathematical model of a three-dimensional reconstruction of a wheel.
[0058] Figure 5 Heat map of wheel tread curvature distribution.
[0059] Table 1 is a data recording table of the simulation experiment of this invention. Detailed Implementation
[0060] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0061] In the description of this application, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.
[0062] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.
[0063] Example 1
[0064] This embodiment provides, for example Figure 1The above describes a TWDS online detection dynamic compensation method with multidimensional pose decoupling, which specifically includes the following steps:
[0065] S1. Deploy multi-dimensional pose sensing units along the track measurement area. The sensing unit includes at least three sets of laser displacement sensors, two sets of inertial measurement units and four induction magnets, used to collect the original two-dimensional contour point cloud of the wheel, the three-axis acceleration and three-axis angular velocity of the sensor bracket, and the wheel through the trigger time stamp, and to calibrate all sensors in a unified spatial coordinate system.
[0066] S2. When the train passes through the measurement area, an accurate timestamp is generated using an induction magnet to synchronize and align the collected raw point cloud frame with the inertial data in time. The instantaneous pose vector of the support is obtained by integrating the inertial data to eliminate the physical vibration deviation of the measurement platform.
[0067] S3. Based on redundant observations from multiple laser sensors and magnetic positioning signals, the composite motion of the wheel relative to the track centerline is decoupled into independent pose components of lateral displacement, yaw angle, and roll angle; a 6-DOF dynamic transformation matrix is constructed for each sampling moment through error components of each dimension.
[0068] S4. The original point cloud is inversely transformed using the dynamic transformation matrix to correct the projection geometric distortion caused by the serpentine swing; the corrected cross-sectional data of multiple frames are spatially fused to establish a high-precision three-dimensional mathematical model of the wheel and its flange area, and key geometric indicators are automatically extracted from the three-dimensional model to detect wheel defects.
[0069] S5. Compare the corrected data measured by multiple sets of sensors to determine the convergence and consistency of the measurement results. When the calculation results exceed the preset industry safety tolerance, a real-time alarm is triggered, and digital maintenance suggestions and evaluation reports are output.
[0070] Preferably, in step S1, a multi-dimensional pose sensing unit is deployed along the track measurement area. The sensing unit includes at least three sets of laser displacement sensors, two sets of inertial measurement units, and four inductive magnets. These are used to acquire the original two-dimensional contour point cloud of the wheel, the three-axis acceleration and three-axis angular velocity of the sensor bracket, and the wheel's trigger timestamp, and to calibrate all sensors using a unified spatial coordinate system. Specifically, this includes the following:
[0071] The three sets of laser displacement sensors are sequentially installed on the outer and inner sides of the track along the rail direction to acquire the original contour point cloud data of the wheel tread and flange area; the two sets of inertial measurement units are installed on the rigid support where the laser displacement sensors are located to measure the three-axis acceleration and three-axis angular velocity of the support in real time when the train passes; the four induction magnets are sequentially installed on the inner side of the track along the rail direction to detect the rail direction position of the wheel center and generate a trigger time stamp signal.
[0072] All sensors are calibrated using a unified spatial coordinate system. The origin is the track centerline, the positive X-axis is along the track direction, the positive Y-axis is perpendicular to the track surface to the left, and the positive Z-axis is perpendicular to the track plane upwards. A world coordinate system is then established with the track centerline as the reference. The local point cloud coordinates of the three sets of laser displacement sensors, the inertial measurement vectors of the two sets of inertial measurement units, and the track orientation information of the magnet are uniformly converted to [the correct coordinates] using a calibration matrix. In the coordinate system, the static system error introduced by the differences in installation position and angle of each sensor is eliminated.
[0073] Preferably, in step S2, when the train passes through the measurement area, a precise timestamp is generated using an induction magnet to synchronize and align the collected raw point cloud frame with the inertial data, and the instantaneous pose vector of the support is obtained through inertial data integration, thus eliminating the physical vibration deviation of the measurement platform; specifically, this includes the following:
[0074] When the train enters the measurement area, the four induction magnets are triggered sequentially, generating a precise time stamp sequence indicating when the wheels reach their preset positions. This generated precise time stamp sequence is recorded as follows: ,in, These represent the moments when the wheelset is detected entering the vehicle and when the protective gate is triggered to open, respectively. This indicates the measurement reference time, corresponding to the moment when the wheel center projection reaches directly below the measurement module. This indicates the moment when the wheelset leaves the measurement area and triggers the protection system to reset. Directly below the measurement module is the ideal measurement position where the laser line of the laser scanner projects onto the wheel tread, corresponding to the wheel center projection point.
[0075] After the system starts up, the sensors work together according to the following process:
[0076] When the train has not arrived, the system is in standby mode, the laser scanner protective gate is closed, and the IMU is in low power mode.
[0077] As the train approaches, magnet number 1 is triggered, waking up the system and activating the cooling system.
[0078] When magnet #2 is triggered, the control unit opens the protective gate of the laser scanner and starts the laser scanner to collect data continuously.
[0079] Magnet No. 3 triggers the measurement reference time. The laser scanner and IMU acquire data synchronously;
[0080] When magnet #4 is triggered, the protective gate closes, and the system returns to standby mode.
[0081] The rising edge of the trigger of the third induction magnet To serve as a synchronization reference, the raw point cloud frames acquired by the laser displacement sensor and the inertial data acquired by the inertial measurement unit are synchronized at the microsecond level.
[0082] The triaxial acceleration and triaxial angular velocity output from the inertial measurement unit are extracted, and the instantaneous spatial displacement vector and deflection vector of the laser displacement sensor bracket when the train passes are obtained through integration.
[0083] By using instantaneous spatial displacement vector and deflection vector, a support vibration compensation matrix is established, and the original two-dimensional contour point cloud is corrected by inverse coordinate transformation. By compensating for the instantaneous displacement and tilt of the support, the support excitation error induced by the high-speed operation of the train is eliminated, and the original wheel point cloud relative to the static track reference frame is obtained.
[0084] Preferably, in step S3, based on redundant observations from multiple laser sensors and magnetic positioning signals, the composite motion of the wheel relative to the track centerline is decoupled into independent pose components of lateral displacement, yaw angle, and roll angle; a 6-DOF dynamic transformation matrix is constructed for each sampling moment using error components of each dimension; specifically, it includes the following:
[0085] By utilizing redundant observation sections formed by three sets of laser displacement sensors along the track direction, combined with the real-time wheel center position determined by induction magnets, a geometric analytical algorithm decouples the complex serpentine oscillation of the wheel relative to the track centerline into independent pose components, including lateral displacement, yaw angle, and roll angle. The lateral displacement is extracted based on the change in lateral distance between the wheel flange and the tread measured by the inner and outer sensors. The yaw angle is calculated by using the observation time difference and distance deviation of the same wheelset side by three sets of scanners distributed along the track direction to determine the wheel's yaw angle around the vertical axis. The roll angle is extracted by analyzing the vertical height difference distribution of tread slope points to determine the wheel's lateral tilt angle relative to the track plane.
[0086] Independent error components in each dimension are obtained by decoupling the lateral swing and head-shaking angle coupling and the vertical jump and side roll motion decoupling, and a 6-DOF dynamic transformation matrix of the scanned point cloud is constructed at each moment.
[0087] The decoupling of lateral sway and head-tilt angle coupling is achieved by using three sets of laser displacement sensors with fixed installation spacing in the rail direction, and comparing the lateral displacement values of the tread obtained by different sensors at the same time. Obtain the pure lateral displacement vector. and yaw angle in radians ,in, This is a correction term for the geometric projection deviation caused by the yaw angle. These represent the lateral displacement observations of the wheel tread obtained by different sensors at the same time, where L is the installation spacing of the sensors along the rail direction;
[0088] The decoupling of vertical bounce and lateral roll motion is achieved by using a point cloud of lateral feature points on the tread acquired by sensors to fit the lateral slope of the tread. And combined with the standard tread slope To distinguish between vertical displacement and tilt angle, the specific formula is as follows: , ,in, R is the observed vertical average height, and R is the nominal radius of the wheel. Let θ be the tilt angle and z be the vertical displacement.
[0089] For the scanned point cloud acquired at each moment, the decoupled independent components are integrated. and limited The pitch angle is calculated by inertial measurement units mounted on a trackside support, which collect the angular velocity around the Y-axis in real time, calculate it through integration, and constrain it to a small value to construct a 6-DOF dynamic transformation matrix at each moment. The dynamic transformation matrix The specific expression is as follows .
[0090] Preferably, in step S4, the original point cloud is subjected to inverse transformation using the dynamic transformation matrix to correct the projection geometric distortion caused by the serpentine oscillation; specifically, this includes:
[0091] Using the 6-DOF dynamic transformation matrix constructed in step S3, an inverse spatial transformation is performed on each frame of the original point cloud to obtain the corrected standard meridional point cloud. ,in, The original point cloud is in homogeneous coordinate form. The coordinates of the standard meridian point cloud are the corrected coordinates. This is the inverse matrix of the dynamic transformation matrix; by calculating and compensating for the yaw, roll, and lateral displacement deviations of the wheel at the moment of measurement, the "pseudo-distorted profile" is restored to a meridional point cloud reflecting the true cross-section of the wheel, such as... Figure 3 As shown, the left side is the original point cloud cross-sectional profile with serpentine distortion before correction, and the right side is the standard meridional point cloud profile after correction by dynamic compensation in step 6-DOF. By comparison, this method effectively eliminates the projection geometric distortion caused by serpentine swaying.
[0092] The step of spatially fusing multi-frame corrected cross-sectional data to establish a high-precision three-dimensional mathematical model of the wheel and its flange region, and automatically extracting key geometric indicators and detecting wheel defects from the three-dimensional model, further includes:
[0093] Multiple frames of corrected two-dimensional cross-sectional data continuously collected during the train's passage are precisely registered and fused according to the spatial position sequence determined by the magnet in step S2. The spatial position sequence of each frame is determined based on the magnet positioning signal, and the spatial position of the i-th frame is represented as... ,in, The starting position is given by v, where v is the wheel speed, calculated by dividing the distance between adjacent magnets by the time difference. The acquisition time of the i-th frame;
[0094] The multi-frame corrected 2D cross-sectional point clouds are stitched together according to their corresponding spatial positions to obtain a fused point cloud set. The fused point cloud is then reconstructed using a moving least squares surface fitting algorithm to generate a continuous and smooth 3D tread surface model. ,like Figure 4 As shown in the figure, the overall three-dimensional shape of the wheel is displayed in an axonometric view, including the rim, tread and rim key areas. In the actual measurement process, through the registration and fusion of multi-frame cross-sectional point clouds, a three-dimensional mathematical model of the wheel containing complete geometric information is reconstructed. This model can accurately reflect the real physical shape of the wheel tread and rim area.
[0095] Key geometric parameters, including rim thickness, rim height, rim slope, and tread wear, are automatically identified and extracted from the 3D model. Wheel defects, such as scratches, pore depth, and pore area, are identified and detected using the variation characteristics of the surface normal vector and curvature of the 3D model. Curvature abrupt change detection technology using 3D point clouds is employed to calculate the average curvature of each point on the model surface. ,in, Let P and P be the maximum and minimum principal curvatures of the surface at point P, respectively. Figure 5 The figure shows a heat map of the curvature distribution of the wheel tread. The curvature is represented by the density of the shaded lines: sparse diagonal lines represent low curvature areas, denser diagonal lines represent medium curvature areas, and dense intersecting lines represent high curvature areas. When the curvature changes significantly at the scratch / porosity defect, it is represented by a high-density shaded area, which can effectively locate the defect.
[0096] And calculate the normal vector of each point p on the surface of the 3D model. The mean curvature Reference curvature of the standard tread area When comparing, and When, point p is marked as a candidate defect point, where, The reference curvature for the healthy tread surface area. Let be the curvature abrupt change threshold, and q be a neighborhood point of p. The threshold for the change in the normal vector;
[0097] Cluster analysis is performed on the marked defect candidate points to segment independent defect regions. Based on the morphological characteristics of the defect regions, scratches and pores are distinguished: those with open boundaries and an area greater than a preset threshold are judged as scratches, and those with closed hole boundaries are judged as pores.
[0098] For the scratched area, calculate the projected area A and the maximum depth. ,when The system immediately detected that the wheel had skid damage and issued a warning. The standard tread profile surface is used; the method for calculating the scratch area A is as follows: the scratch area point set is triangulated, and the total area is obtained by summing the areas of each triangle;
[0099] For the pore region, identify the defect region with closed pore boundary, calculate the normal distance from the deepest point in the pore to the standard tread profile as the pore depth, and triangulate the pore opening region to obtain the pore area.
[0100] The maximum depth The calculation method is as follows: calculate the normal distance from each point in the scratch area to the fitted surface of the standard tread profile, and take the maximum value.
[0101] Preferably, in step S5, when the calculation result exceeds the preset industry safety tolerance, an automatic real-time alarm is triggered at the red warning level. The measured three-dimensional profile is compared with the standard template profile, and the minimum overhaul depth of the wheelset under the premise of meeting safety standards is automatically calculated. Digital maintenance suggestions and vehicle operating status assessment reports are output, specifically including the following:
[0102] Mutual inductance verification is performed on the geometric parameters of the three sets of laser displacement sensors after independent calculation and correction to determine the convergence of the measurement results. When the data deviation between each group is within the preset threshold, the weighted average value is taken as the final measured value to ensure the stability and reliability of the output data.
[0103] The extracted key geometric indicators are compared with the preset industry safety tolerances in real time. The safety tolerances include the upper limit of the flange height, the lower limit of the flange thickness, and the upper limit of the tread wear. When the calculated result of any indicator exceeds the preset industry safety upper limit, the system automatically determines it to be a red warning level and triggers a real-time alarm signal to notify the dispatching department to carry out vehicle impoundment inspection.
[0104] The measured 3D profile reconstructed by step S4 Contour of standard template Spatial registration and interpolation are performed, and the globally optimal cutting path is found through an iterative algorithm. The minimum overhaul depth of the wheel is automatically calculated. ,in, The target outline after rotation modification;
[0105] It automatically summarizes measurement results, including: rim height, rim thickness, tread wear, scratch detection results, and minimum repair depth, and outputs a digital assessment report including minimum repair recommendations, expected post-repair life, and overall vehicle operating safety status to guide repair workshops in making precise repairs.
[0106] Example 2
[0107] This embodiment also provides a TWDS online detection dynamic compensation system with multi-dimensional pose decoupling, specifically including a multi-dimensional pose perception unit, a time synchronization and compensation unit, a multi-dimensional pose decoupling unit, a three-dimensional reconstruction unit, a defect detection unit, and an early warning and decision output unit;
[0108] The multidimensional pose sensing unit consists of at least three sets of laser displacement sensors, two sets of inertial measurement units, and four inductive magnets. It is deployed along the track measurement area to collect the original two-dimensional contour point cloud of the wheel, the three-axis acceleration and three-axis angular velocity of the sensor bracket, and the trigger timestamp of the wheel. All sensors in the sensing unit are calibrated using a unified spatial coordinate system.
[0109] The time synchronization and compensation unit is electrically connected to the multi-dimensional pose sensing unit. It is used to generate a precise timestamp using an inductive magnet, synchronize and align the collected original point cloud frame with the inertial data, and obtain the instantaneous pose vector of the support through inertial data integration to eliminate the physical vibration deviation of the measurement platform.
[0110] The multi-dimensional pose decoupling unit is connected to the time synchronization and compensation unit. It is used to decouple the composite motion of the wheel relative to the center line of the track into independent pose components of lateral displacement, yaw angle and roll angle based on redundant observations from multiple sets of laser sensors and magnetic positioning signals, and to construct a 6-DOF dynamic transformation matrix for each sampling time.
[0111] The three-dimensional reconstruction unit is connected to the multi-dimensional pose decoupling unit and is used to perform inverse transformation processing on the original point cloud using the dynamic transformation matrix, correct the projection geometric distortion caused by the serpentine swing, and spatially fuse the corrected cross-sectional data of multiple frames to establish a high-precision three-dimensional mathematical model of the wheel and its flange area.
[0112] The defect detection unit is connected to the three-dimensional reconstruction unit and is used to automatically extract key geometric indicators from the three-dimensional model and detect wheel defects.
[0113] The early warning and decision output unit is connected to the defect detection unit and is used to compare the corrected data measured by multiple sets of sensors to determine the convergence and consistency of the measurement results. When the calculation results exceed the preset industry safety tolerance, a real-time alarm is triggered, and digital maintenance suggestions and evaluation reports are output.
[0114] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0115] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the TWDS online detection dynamic compensation method for multi-dimensional pose decoupling as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0116] Example 3
[0117] The following is another embodiment of the present invention, which provides a TWDS online detection dynamic compensation method and system with multidimensional pose decoupling. In order to verify the beneficial effects of the present invention, a simulation experiment is conducted for scientific demonstration.
[0118] This experiment aims to verify the effectiveness of a multi-dimensional pose decoupling TWDS online detection dynamic compensation method. Through multi-dimensional pose perception, time synchronization alignment, composite motion decoupling, inverse transformation distortion correction, and 3D modeling, it improves the detection accuracy and robustness of a freight car fault dynamic detection system under complex operating conditions. The experiment uses simulated train passing data and raw signals collected on-site, including the vibration acceleration of the sensor bracket, wheel swaying pose, original 2D point cloud cross-sections, and magnet trigger timescales. By analyzing the deviations of the measured wheel flange geometric dimensions before and after compensation from the standard values, the accuracy of the system in eliminating bracket vibration and wheel pose distortion is verified.
[0119] The simulation experiment steps are implemented according to the content of the TWDS online detection dynamic compensation method and system for multi-dimensional pose decoupling provided in Example 1, and the specific steps include:
[0120] Multidimensional pose sensing units are deployed along the track measurement area, including at least three sets of laser displacement sensors, two sets of inertial measurement units and four inductive magnets. A world coordinate system with the track centerline as the reference is established, and all sensors are spatially calibrated.
[0121] When the simulated train passes through the measurement area, the trigger timestamp sequence generated by the induction magnet is used to synchronize and align the collected high-frequency raw point cloud frames with the IMU inertial data at the microsecond level.
[0122] By performing real-time integration on the inertial data, the instantaneous pose vector of the sensor bracket at the moment the train passes is obtained, thus eliminating the vibration deviation of the measurement platform caused by strong impact at the physical level.
[0123] By utilizing the redundant observation sections of three sets of laser sensors and combining them with the magnetic positioning signal, the composite motion of the wheel relative to the centerline of the track is decoupled into independent components of lateral displacement, yaw angle, and roll angle, and a 6-DOF dynamic transformation matrix is constructed for each sampling moment.
[0124] The original two-dimensional point cloud is inversely transformed using a dynamic transformation matrix to correct the projection geometric distortion caused by the swaying of the wheel. The corrected cross-sectional data from multiple frames are spatially fused to reconstruct a high-precision three-dimensional mathematical model of the wheel and flange area. Key geometric indicators such as flange thickness and flange height are automatically extracted, and defect detection and safety tolerance comparison are performed.
[0125] The specific data from the above simulation experiment are as follows:
[0126]
[0127] Table 1
[0128] Experimental Analysis:
[0129] By comparing the original distorted point cloud in S1 with the standard cross-sectional data corrected in S4, the superior performance of this invention in 6-DOF dynamic compensation was verified. Experimental data showed that when the bracket generates a large instantaneous pose vector and the wheel has significant yaw, the system can accurately extract the wheel diameter and identify minor defects such as tread abrasion and wheel flange peeling. The consistency judgment and maintenance suggestions output in step S5 are completely consistent with the actual simulation state, proving the effectiveness of this method in improving the accuracy of TWDS online detection and guiding digital maintenance.
[0130] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.
Claims
1. A TWDS online detection and dynamic compensation method with multidimensional pose decoupling, characterized in that: Includes the following steps: S1. Deploy multi-dimensional pose sensing units along the track measurement area. The multi-dimensional pose sensing unit includes at least three sets of laser displacement sensors, two sets of inertial measurement units and four induction magnets, used to collect the original two-dimensional contour point cloud of the wheel, the three-axis acceleration and three-axis angular velocity of the sensor bracket, and the wheel through the trigger time stamp, and to perform unified spatial coordinate system calibration on all sensors. S2. When the train passes through the measurement area, an accurate timestamp is generated using an induction magnet to synchronize and align the collected raw point cloud frame with the inertial data in time. The instantaneous pose vector of the support is obtained by integrating the inertial data to eliminate the physical vibration deviation of the measurement platform. S3. Based on redundant observations from multiple laser sensors and magnetic positioning signals, the composite motion of the wheel relative to the track centerline is decoupled into independent pose components of lateral displacement, yaw angle, and roll angle; a 6-DOF dynamic transformation matrix is constructed for each sampling moment through error components of each dimension. S4. The original point cloud is inversely transformed using the dynamic transformation matrix to correct the projection geometric distortion caused by the serpentine swing; the corrected cross-sectional data of multiple frames are spatially fused to establish a high-precision three-dimensional mathematical model of the wheel and its flange area, and key geometric indicators are automatically extracted from the three-dimensional model to detect wheel defects. S5. Compare the corrected data measured by multiple sets of sensors to determine the convergence and consistency of the measurement results. When the calculation results exceed the preset industry safety tolerance, a real-time alarm is triggered, and digital maintenance suggestions and evaluation reports are output.
2. The TWDS online detection dynamic compensation method for multidimensional pose decoupling according to claim 1, characterized in that: In step S1, the three sets of laser displacement sensors are sequentially installed on the outer and inner sides of the track along the rail direction to acquire the original contour point cloud data of the wheel tread and flange area; the two sets of inertial measurement units are installed on the rigid support where the laser displacement sensors are located to measure the three-axis acceleration and three-axis angular velocity of the support in real time when the train passes; the four induction magnets are sequentially installed on the inner side of the track along the rail direction to detect the rail direction position of the wheel center and generate a trigger time stamp signal. All sensors are calibrated using a unified spatial coordinate system. The origin is the track centerline, the positive X-axis is along the track direction, the positive Y-axis is perpendicular to the track surface to the left, and the positive Z-axis is perpendicular to the track plane upwards. A world coordinate system is then established with the track centerline as the reference. The local point cloud coordinates of the three sets of laser displacement sensors, the inertial measurement vectors of the two sets of inertial measurement units, and the track orientation information of the magnet are uniformly converted to [the correct coordinates] using a calibration matrix. In the coordinate system, the static system error introduced by the differences in installation position and angle of each sensor is eliminated.
3. The TWDS online detection dynamic compensation method for multidimensional pose decoupling according to claim 1, characterized in that: In step S2, when the train enters the measurement area, the four induction magnets are triggered sequentially to generate a precise timestamp sequence of the wheels reaching each preset position. This precise timestamp sequence is recorded as follows: ,in, These represent the moments when the wheelset is detected entering the vehicle and when the protective gate is triggered to open, respectively. Indicates the reference time for measurement. This indicates the moment when the wheelset leaves the measurement area and triggers the protection system to reset; The rising edge of the trigger of the third induction magnet To serve as a synchronization reference, the raw point cloud frames acquired by the laser displacement sensor and the inertial data acquired by the inertial measurement unit are synchronized at the microsecond level. The triaxial acceleration and triaxial angular velocity output from the inertial measurement unit are extracted, and the instantaneous spatial displacement vector and deflection vector of the laser displacement sensor bracket when the train passes are obtained through integration. By using instantaneous spatial displacement vector and deflection vector, a support vibration compensation matrix is established, and the original two-dimensional contour point cloud is corrected by inverse coordinate transformation; by compensating for the instantaneous displacement and tilt of the support, the original wheel point cloud relative to the static track reference frame is obtained.
4. The TWDS online detection dynamic compensation method for multidimensional pose decoupling according to claim 1, characterized in that: In step S3, redundant observation sections formed by three sets of laser displacement sensors along the track direction are used, combined with the real-time position of the wheel center determined by the induction magnet. A geometric analytical algorithm is used to decouple the complex serpentine oscillation of the wheel relative to the track centerline into independent pose components, including lateral displacement, yaw angle, and roll angle. The lateral displacement is extracted based on the change in the lateral distance between the wheel flange and the tread measured by the inner and outer sensors. The yaw angle is calculated by using the observation time difference and distance deviation of the same wheelset side by three sets of scanners distributed along the track direction to calculate the wheel's yaw angle around the vertical axis. The roll angle is extracted by analyzing the height difference distribution of the tread slope points in the vertical direction to extract the lateral tilt angle of the wheel relative to the track plane. Independent error components in each dimension are obtained by decoupling the lateral swing and head-shaking angle coupling and the vertical jump and side roll motion decoupling, and a 6-DOF dynamic transformation matrix of the scanned point cloud is constructed at each moment. The decoupling of lateral sway and head-tilt angle coupling is achieved by using three sets of laser displacement sensors with fixed installation spacing in the rail direction, and comparing the lateral displacement values of the tread obtained by different sensors at the same time. Obtain the pure lateral displacement vector. and yaw angle in radians ,in, This is a correction term for the geometric projection deviation caused by the yaw angle. These represent the lateral displacement observations of the wheel tread obtained by different sensors at the same time, where L is the installation spacing of the sensors along the rail direction; The decoupling of vertical bounce and lateral roll motion is achieved by using a point cloud of lateral feature points on the tread acquired by sensors to fit the lateral slope of the tread. And combined with the standard tread slope To distinguish between vertical displacement and tilt angle, the specific formula is as follows: , ,in, R is the observed vertical average height, and R is the nominal radius of the wheel. Let θ be the tilt angle and z be the vertical displacement. For the scanned point cloud acquired at each moment, the decoupled independent components are integrated. and limited The pitch angle is calculated by integrating the angular velocity around the Y-axis in real time using an inertial measurement unit mounted on a support bracket beside the track, thus constructing a 6-DOF dynamic transformation matrix for each moment. .
5. The TWDS online detection dynamic compensation method for multidimensional pose decoupling according to claim 1, characterized in that: In step S4, the original point cloud is inversely transformed using the dynamic transformation matrix to correct the projection geometric distortion caused by the serpentine oscillation; specifically, this includes: Using the 6-DOF dynamic transformation matrix constructed in step S3, a spatial inverse transformation is performed on each frame of the original point cloud to obtain the corrected standard meridional point cloud. ,in, The original point cloud is in homogeneous coordinate form. The coordinates of the standard meridian point cloud are as follows; It is the inverse of the dynamic transformation matrix.
6. The TWDS online detection dynamic compensation method for multidimensional pose decoupling according to claim 1, characterized in that: In step S4, the step of spatially fusing the corrected cross-sectional data from multiple frames to establish a high-precision three-dimensional mathematical model of the wheel and its flange region, and automatically extracting key geometric indicators from the three-dimensional model and detecting wheel defects includes: Multiple frames of corrected two-dimensional cross-sectional data continuously collected during the train's passage are precisely registered and fused according to the spatial position sequence determined by the magnet in step S2. The spatial position sequence of each frame is determined based on the magnet positioning signal, and the spatial position of the i-th frame is represented as... ,in, Here is the starting position, and v is the wheel speed. The acquisition time of the i-th frame; The multi-frame corrected 2D cross-sectional point clouds are stitched together according to their corresponding spatial positions to obtain a fused point cloud set. The fused point cloud is then reconstructed using a moving least squares surface fitting algorithm to generate a continuous and smooth 3D tread surface model. ; Key geometric parameters, including rim thickness, rim height, rim slope, and tread wear, are automatically identified and extracted from the 3D model. Wheel defects, such as scratches, pore depth, and pore area, are identified and detected using the variation characteristics of the surface normal vector and curvature of the 3D model. Curvature abrupt change detection technology using 3D point clouds is employed to calculate the average curvature of each point on the model surface. ,in, These are the maximum and minimum principal curvatures of the surface at point p, respectively; And calculate the normal vector of each point p on the surface of the 3D model. The mean curvature Reference curvature of the standard tread area When comparing, and When, point p is marked as a candidate defect point, where, The reference curvature for the healthy tread surface area. Let be the curvature abrupt change threshold, and q be a neighborhood point of p. The threshold for the change in the normal vector; Cluster analysis is performed on the marked defect candidate points to segment independent defect regions. Based on the morphological characteristics of the defect regions, scratches and pores are distinguished: those with open boundaries and an area greater than a preset threshold are judged as scratches, and those with closed hole boundaries are judged as pores. For the scratched area, calculate the projected area A and the maximum depth. ,when The system detects wheel slippage and issues a warning. The standard tread profile surface is used; the method for calculating the scratch area A is as follows: the scratch area point set is triangulated, and the total area is obtained by summing the areas of each triangle; For the pore region, identify the defect region with closed pore boundary, calculate the normal distance from the deepest point in the pore to the standard tread profile as the pore depth, and triangulate the pore opening region to obtain the pore area. The maximum depth The calculation method is as follows: calculate the normal distance from each point in the scratch area to the fitted surface of the standard tread profile, and take the maximum value.
7. The TWDS online detection dynamic compensation method for multidimensional pose decoupling according to claim 1, characterized in that: In step S5, when the calculation result exceeds the preset industry safety tolerance, an automatic real-time alarm is triggered at the red warning level. The measured three-dimensional profile is compared with the standard template profile, and the minimum overhaul depth of the wheelset is automatically calculated under the premise of meeting safety standards. Digital maintenance suggestions and vehicle operating status assessment reports are output, including the following: Mutual inductance verification is performed on the geometric parameters of the three sets of laser displacement sensors after independent calculation and correction to determine the convergence of the measurement results. When the data deviation between each group is within the preset threshold, the weighted average value is taken as the final measured value. The extracted key geometric indicators are compared with the preset industry safety tolerances in real time. The safety tolerances include the upper limit of the flange height, the lower limit of the flange thickness, and the upper limit of the tread wear. When the calculated result of any indicator exceeds the preset industry safety upper limit, the system automatically determines it to be a red warning level and triggers a real-time alarm signal to notify the dispatching department to carry out vehicle impoundment inspection. The measured 3D profile reconstructed in step S4 Contour of standard template Spatial registration and interpolation are performed, and the globally optimal cutting path is found through an iterative algorithm. The minimum overhaul depth of the wheel is automatically calculated. ,in, The target outline after rotation modification; It automatically summarizes measurement results, including: rim height, rim thickness, tread wear, scratch detection results, and minimum repair depth, and outputs a digital assessment report including minimum repair recommendations, expected post-repair life, and overall vehicle operating safety status to guide repair workshops in making precise repairs.
8. The TWDS online detection dynamic compensation system for multidimensional pose decoupling according to claim 1, characterized in that: It includes a multi-dimensional pose perception unit, a time synchronization and compensation unit, a multi-dimensional pose decoupling unit, a three-dimensional reconstruction unit, a defect detection unit, and an early warning and decision output unit; The multidimensional pose sensing unit consists of at least three sets of laser displacement sensors, two sets of inertial measurement units, and four inductive magnets. It is deployed along the track measurement area to collect the original two-dimensional contour point cloud of the wheel, the three-axis acceleration and three-axis angular velocity of the sensor bracket, and the trigger timestamp of the wheel. All sensors in the sensing unit are calibrated using a unified spatial coordinate system. The time synchronization and compensation unit is electrically connected to the multi-dimensional pose sensing unit. It is used to generate a precise timestamp using an inductive magnet, synchronize and align the collected original point cloud frame with the inertial data, and obtain the instantaneous pose vector of the support through inertial data integration to eliminate the physical vibration deviation of the measurement platform. The multi-dimensional pose decoupling unit is connected to the time synchronization and compensation unit. It is used to decouple the composite motion of the wheel relative to the center line of the track into independent pose components of lateral displacement, yaw angle and roll angle based on redundant observations from multiple sets of laser sensors and magnetic positioning signals, and to construct a 6-DOF dynamic transformation matrix for each sampling time. The three-dimensional reconstruction unit is connected to the multi-dimensional pose decoupling unit and is used to perform inverse transformation processing on the original point cloud using the dynamic transformation matrix, correct the projection geometric distortion caused by the serpentine swing, and spatially fuse the corrected cross-sectional data of multiple frames to establish a high-precision three-dimensional mathematical model of the wheel and its flange area. The defect detection unit is connected to the three-dimensional reconstruction unit and is used to automatically extract key geometric indicators from the three-dimensional model and detect wheel defects. The early warning and decision output unit is connected to the defect detection unit and is used to compare the corrected data measured by multiple sets of sensors to determine the convergence and consistency of the measurement results. When the calculation results exceed the preset industry safety tolerance, a real-time alarm is triggered, and digital maintenance suggestions and evaluation reports are output.