Intelligent collaborative rail transit depot automated maintenance assembly line system

Abstract

This invention belongs to the field of data processing technology, specifically relating to an intelligent collaborative automated maintenance line system for rail transit vehicle depots. It includes: a multi-source data synchronous acquisition module, a dynamic trajectory calculation module, a spatiotemporal resampling mapping module, and an alignment quality self-check output module. The system synchronously acquires motion-driven data, constraint anchor point data, and monitoring load data. The motion-driven data and constraint anchor point data are input into a dynamic model. The absolute coordinates of the anchor points are used to flexibly correct the projected trajectory, calculate the virtual correction acceleration, update the corrected position, and construct a spatiotemporal mapped trajectory sequence. The sequence undergoes monotonicity processing, and the monitoring load data is mapped to spatial grid points to generate spatially aligned monitoring data. A quality self-check is performed based on the alignment confidence residual. This invention effectively solves the problem of multi-source data alignment under unsteady train operating conditions, ensuring data physical continuity and improving the accuracy and reliability of fault diagnosis.

Description

Technical Field

[0001] This invention relates to the field of data processing technology. More specifically, this invention relates to an intelligent and collaborative automated maintenance line system for rail transit vehicle depots. Background Technology

[0002] In the field of intelligent operation and maintenance of rail transit vehicles, various sensors such as linear array cameras, acoustic arrays, and infrared probes are usually arranged longitudinally along the track on the vehicle entry inspection line in the depot. These sensors are used to collect the health status of the train as it passes by. Since different sensors have fixed installation spacing in physical space, the time for collecting data on the same physical component is different. Therefore, it is necessary to perform spatiotemporal alignment of multi-source heterogeneous monitoring data and map the scattered data onto the same physical coordinates of the train body to achieve multimodal fault fusion diagnosis.

[0003] Existing technologies for solving such spatiotemporal alignment problems generally employ a geometric segmentation correction method based on odometer calculations. This method mainly relies on a speed measuring device to obtain train speed and integrate it to estimate the travel distance. It also uses axle-counting magnets laid at fixed intervals along the track as absolute position anchor points. When the system detects that the train has passed an anchor point, it calculates the deviation between the integrated mileage and the absolute coordinates of the anchor point. Then, it uses linear interpolation or segmented scaling to distribute the deviation evenly to the previous travel trajectory, attempting to eliminate accumulated errors through geometric forced stretching or compression in order to achieve the correspondence between data and physical position.

[0004] However, this correction method based on linear geometric assumptions has revealed serious adaptability defects in actual engineering. The scenario of trains entering the depot is extremely complex. Trains often experience non-steady-state conditions such as low-speed creep, frequent starts and stops, and even slight slippage. In these dynamic processes, the motion characteristics of the train exhibit a high degree of nonlinearity. The existing linear interpolation algorithm cannot truly reflect the complex speed fluctuations of the train between anchor points. Especially when wheel slippage causes severe integral drift, or when the train stops midway and disrupts the time continuity, simple geometric stretching not only fails to eliminate errors but also introduces false position jumps or signal distortions. This directly leads to significant misalignment of data from different sensors in spatial alignment, seriously interfering with the subsequent diagnostic system's extraction and identification of minor fault features. Summary of the Invention

[0005] To address the technical problem that existing linear geometric correction methods are ill-suited to unsteady train conditions, leading to spatial misalignment of monitoring data and affecting diagnostic reliability, this invention provides an intelligent collaborative automated maintenance line system for rail transit vehicle depots. The system includes the following modules:

[0006] The multi-source data synchronous acquisition module is used to synchronously acquire motion drive data, constraint anchor point data, and monitoring load data, and align all acquired data to a unified time base. The dynamic trajectory calculation module is used to input motion drive data and constraint anchor point data into a pre-built dynamic model, use the absolute coordinates of anchor points in the constraint anchor point data to perform flexible correction on the trajectory derived from the motion drive data, calculate the virtual correction acceleration at the current moment, update the corrected velocity at the current moment based on the virtual correction acceleration, and then update the corrected position at the current moment, forming a spatiotemporal mapped trajectory sequence at each moment. The spatiotemporal resampling mapping module is used to perform monotonicity processing on the spatiotemporal mapped trajectory sequence, and based on the monotonic spatiotemporal mapped trajectory sequence, map the monitoring load data to spatial grid points in the train body coordinate system to generate spatial alignment monitoring data. The alignment quality self-check output module is used to calculate the alignment confidence residual during the train's passage based on the virtual correction acceleration and corrected velocity at each moment, and output spatial alignment monitoring data when the alignment confidence residual is less than or equal to a preset safety threshold.

[0007] This invention constructs a dynamic model incorporating constrained anchor point data and utilizes virtual correction acceleration to flexibly correct trajectories derived from motion-driven data. This smoothly eliminates accumulated velocity integral errors while avoiding positional jumps at anchor points, a problem inherent in traditional rigid geometric correction. This ensures the physical continuity and accuracy of the spatiotemporal mapped trajectory sequence under unsteady train motion. Furthermore, by applying monotonicity processing to the spatiotemporal mapped trajectory sequence and combining it with spatial grid point mapping technology, this invention mounts discrete and heterogeneous monitoring load data in the time domain to a unified train body coordinate system, effectively overcoming data spatial distortion and misalignment problems caused by train acceleration, deceleration, or slippage. Finally, by calculating alignment confidence residuals, this invention establishes a data quality self-checking mechanism that automatically identifies and blocks abnormal alignment results that violate physical laws, thereby improving the accuracy and diagnostic reliability of multi-source data fusion in complex operating conditions of the automated maintenance system for rail transit depots.

[0008] Preferably, the motion drive data includes the instantaneous speed of the train; the constraint anchor point data includes the trigger time of the anchor point and the absolute coordinates of the anchor point; and the monitoring load data includes the raw monitoring data output by the trackside monitoring subsystems installed at different physical locations.

[0009] Preferably, the method for obtaining the anchor point is as follows: an axle-counting magnet installed at fixed intervals along the track monitors the pulse signal passing by the wheel. When the rising edge of the pulse is detected, the trigger time is recorded, and the absolute coordinates are obtained according to the preset physical installation position of the magnet. The trigger time and the absolute coordinates are combined to define the anchor point.

[0010] Preferably, calculating the virtual correction acceleration at the current moment includes: determining whether the absolute value of the time difference between the current sampling moment and the triggering moment of the nearest anchor point is less than the anchor triggering time window threshold; in response to the absolute value of the time difference being less than the anchor triggering time window threshold, determining the product of the square of the system's natural frequency parameter and the position deviation as the virtual correction acceleration at the current moment; otherwise, setting the virtual correction acceleration at the current moment to 0; wherein, the position deviation is the difference between the absolute coordinates of the nearest anchor point at the current sampling moment and the corrected position at the previous moment.

[0011] This invention controls the timing of virtual correction acceleration by setting a time window threshold based on anchor point triggering. Only when the train passes through the effective physical area near the constraint anchor point is a flexible traction force proportional to the position deviation introduced, while at other times it remains in a free motion state. By utilizing the physical properties of Hooke's Law, the invention eliminates the cumulative error of speed integral while avoiding the trajectory jump caused by traditional forced correction, thus ensuring a smooth transition of the train's trajectory in time and space.

[0012] Preferably, updating the corrected speed at the current moment based on the virtual correction acceleration includes: calculating the difference between the corrected speed at the previous moment and the instantaneous speed of the train at the current moment, as a speed error term; calculating twice the product of the system's natural frequency parameter, damping ratio, and the speed error term, as a damping adjustment term; calculating the difference between the virtual correction acceleration at the current moment and the damping adjustment term, multiplying the result by the time step to obtain the speed correction amount; and adding the corrected speed at the previous moment to the speed correction amount to obtain the corrected speed at the current moment.

[0013] This invention introduces a damping adjustment mechanism that includes a speed error term. While using virtual correction acceleration to eliminate position deviation, the damping term is used to dynamically adjust the corrected speed. By calculating the difference between the corrected speed at the previous moment and the current instantaneous speed and applying damping, the system can force the corrected speed curve to closely follow the changing trend of the train's actual instantaneous speed. At the same time, the dissipation characteristics of damping are used to effectively filter out high-frequency jitter noise in the speed measurement data, prevent non-physical oscillations caused by simple position traction, and ensure the convergence and stability of the speed correction process.

[0014] Preferably, updating the corrected position at the current moment includes: calculating the product of the corrected velocity at the current moment and the time step as the position increment; adding the corrected position at the previous moment to the position increment to obtain the corrected position at the current moment.

[0015] Preferably, the monotonicity processing of the spatiotemporal mapping trajectory sequence includes: setting an initial value for the monotonic correction position, traversing the spatiotemporal mapping trajectory sequence, and setting the monotonic correction position of the current moment to the monotonic correction position of the previous moment in response to the current moment's corrected position being less than the previous moment's monotonic correction position; and setting the current moment's monotonic correction position to the current moment's corrected position in response to the current moment's corrected position being greater than or equal to the previous moment's monotonic correction position.

[0016] This invention performs monotonicity processing on the spatiotemporal mapped trajectory sequence, forcibly filtering out non-physical backtracking phenomena that may occur in the trajectory. This ensures that the calculated position of the train during the entry process is strictly non-decreasing over time, thereby guaranteeing that the position-to-time mapping function has a unique inverse function. This effectively solves the problem of slight trajectory reversal caused by sensor noise or calculation errors, ensuring that a unique and accurate timestamp can be obtained when looking up the time based on the position, and avoiding ambiguity in data mapping.

[0017] Preferably, mapping the monitoring load data to spatial grid points in the train body coordinate system includes: constructing a spatial grid point sequence with the train head as the origin; for each trackside monitoring subsystem that outputs monitoring load data, adding the fixed installation coordinates of the trackside monitoring subsystem to the relative coordinate values ​​of each spatial grid point in the spatial grid point sequence to obtain the target absolute track coordinates for each spatial grid point of the trackside monitoring subsystem.

[0018] This invention achieves a precise conversion from the vehicle coordinate system to the track coordinate system by constructing a unified spatial grid point sequence for the train body and calculating the corresponding absolute track coordinate target value by combining the fixed installation coordinates of each trackside monitoring subsystem. It clarifies the absolute position of the train when different sensors collect the same physical slice of the vehicle body, effectively compensates for the spatial phase difference caused by the different physical installation spacing of the sensors, and provides an accurate spatial index for uniformly mounting dispersed and heterogeneous monitoring data to the train body coordinate system.

[0019] Preferably, the generation of spatial alignment monitoring data includes: for each track absolute coordinate target value, searching in the spatiotemporal mapping trajectory sequence after monotonicity processing to determine the sampling time index that makes the monotonic correction position fall within the position interval containing the track absolute coordinate target value; based on the sampling time index, using linear interpolation to calculate the precise time when the train arrives at the track absolute coordinate target value; and resampling in the monitoring load data according to the precise time to obtain spatial alignment monitoring data that corresponds one-to-one with the spatial grid point sequence.

[0020] This invention employs a reverse search and linear interpolation technique based on monotonic trajectories to achieve a high-precision inverse mapping from spatial position to precise time. By determining the precise time when the train arrives at each track absolute coordinate target value and resampling the monitoring load data accordingly, it eliminates the scaling and distortion of data in the time domain caused by unsteady motions such as train speed changes and crawling. This ensures that the resampled spatially aligned monitoring data is strictly aligned in physical space, enabling different sensors to accurately describe the health status of the same geometric position on the train body.

[0021] Preferably, the step of calculating the alignment confidence residual during the train's passage based on the virtual corrected acceleration and corrected speed at each moment includes: calculating the absolute value of the product of the virtual corrected acceleration, corrected speed, and time step at each moment as the virtual work done at that moment; and summing up the virtual work done at all moments during the train's passage to obtain the alignment confidence residual during the train's passage.

[0022] This invention utilizes the energy and work principles in a dynamic model to construct an alignment confidence residual as a self-checking index. By calculating and accumulating the virtual work done by the virtual correction acceleration during the correction process, the degree of conflict between the input data and the physical model can be evaluated. When the input speed is severely distorted due to wheel slippage or sensor failure, the model needs to do a huge amount of work to forcibly pull back the anchor point, resulting in a high residual. Based on this, the invention can automatically identify and eliminate unreliable alignment results, effectively preventing abnormal data from flowing into the downstream diagnostic process and ensuring the robustness and security of the system.

[0023] The beneficial effects of this invention are as follows:

[0024] This invention establishes a unified time reference through a multi-source data synchronous acquisition module and introduces a dynamic trajectory calculation module based on physical laws. It uses the absolute coordinates of constraint anchor point data to generate virtual correction acceleration, performing flexible correction on the trajectory derived from motion-driven data. This eliminates velocity integral accumulation errors while avoiding position jumps caused by traditional rigid geometric correction, ensuring the smoothness and physical authenticity of the spatiotemporal mapped trajectory sequence. Furthermore, this invention uses a spatiotemporal resampling mapping module to remove non-physical backtracking noise from the trajectory and accurately maps the time-domain discrete monitoring load data to spatial grid points in the train body coordinate system. This effectively eliminates spatiotemporal distortions caused by unsteady motions such as train deceleration and slippage, achieving strict alignment of multi-source heterogeneous data in physical space. Finally, this invention utilizes an alignment quality self-check output module to construct a closed-loop verification mechanism. Based on the principle of dynamic work, it calculates the alignment confidence residual, automatically identifies and intercepts abnormal data, thereby improving the overall alignment accuracy and diagnostic reliability of the rail transit automated maintenance system under complex operating conditions. Attached Figure Description

[0025] Figure 1 This is a schematic diagram illustrating the intelligent collaborative automated maintenance assembly line system for rail transit vehicle depots in this invention.

[0026] Figure 2 This is a schematic diagram of the time-domain waveform of the original monitoring data;

[0027] Figure 3 This is a schematic diagram illustrating the change of virtual correction acceleration over time.

[0028] Figure 4 This is a schematic diagram comparing the traditional rigid geometric correction trajectory, the spatiotemporal mapping trajectory sequence of this invention, and the real physical trajectory.

[0029] Figure 5 This is a waveform diagram of spatial alignment monitoring data. Detailed Implementation

[0030] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] The specific embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0032] This invention provides an intelligent and collaborative automated maintenance line system for rail transit vehicle depots. For example... Figure 1 As shown, the intelligent collaborative automated maintenance line system for rail transit vehicle depots includes a multi-source data synchronous acquisition module 100, a dynamic trajectory calculation module 200, a spatiotemporal resampling mapping module 300, and an alignment quality self-check output module 400, which are described in detail below.

[0033] The multi-source data synchronous acquisition module 100 is used to synchronously acquire motion drive data, constraint anchor point data and monitoring load data, and align all acquired data to a unified time base.

[0034] It should be noted that due to frequent acceleration and deceleration and slight slippage during the train's entry into the depot, relying solely on speed integrals will result in accumulated errors that diverge over time. Furthermore, relying solely on discrete axle counting signals cannot describe the nonlinear motion details between the two anchor points, leading to misalignment in subsequent data alignment. Therefore, this invention employs a multi-source fusion strategy, simultaneously collecting high-frequency motion drive data, constraint anchor point data, and monitoring load data to establish a unified time reference, providing driving and constraint forces for the subsequent construction of the dynamic model.

[0035] Specifically, motion drive data, constraint anchor point data, and monitoring load data are synchronously collected at a preset sampling frequency. The motion drive data includes the train's instantaneous speed, acquired via a speed-measuring radar or onboard wheel axle encoder installed along the trackside. The constraint anchor point data is acquired by using axle-counting magnets installed at fixed intervals along the track to monitor the pulse signals passing by the wheels. When the rising edge of the pulse is detected, the trigger time is recorded, and the absolute coordinates are obtained based on the preset physical installation position of the magnets. The combination of the trigger time and the absolute coordinates is defined as the anchor point. The monitoring load data includes raw monitoring data output from different trackside monitoring subsystems installed at different physical locations, such as thermistor voltage signals output from the infrared axle temperature detection subsystem and vibration sound pressure signals output from the trackside acoustic diagnostic subsystem. All collected data is aligned to a unified NTP timestamp. For example, Figure 2 The diagram shows the time-domain waveform of the original monitoring data. It can be seen that the original monitoring data output by different trackside monitoring subsystems are seriously misaligned in the time domain.

[0036] It should be further explained that, regarding the sampling frequency, considering that the numerical solution of the second-order differential equation relies on high-density time slices to smoothly simulate the action of transient forces, in order to ensure that the dynamic model can perform a sufficient number of calculus iterations within the extremely short anchor trigger time window, so that the subsequent virtual correction acceleration exhibits continuous and gradual physical characteristics rather than discrete pulse interference, this embodiment sets the sampling frequency to... time step for If the sampling frequency is too low, the time window for subsequent virtual correction acceleration will be too short, causing numerical oscillations. If the sampling frequency is too high, it will increase the computational burden on the processor. In other embodiments, the implementer can set the sampling frequency according to the processor's computing power and the requirements for trajectory smoothness.

[0037] The dynamic trajectory calculation module 200 is used to input motion-driven data and constraint anchor point data into a pre-constructed dynamic model, use the absolute coordinates of the anchor points in the constraint anchor point data to perform flexible correction on the trajectory derived from the motion-driven data, calculate the virtual correction acceleration at the current moment, update the corrected velocity at the current moment according to the virtual correction acceleration, and then update the corrected position at the current moment, and construct a spatiotemporal mapped trajectory sequence with the corrected positions at each moment.

[0038] It should be noted that traditional rigid geometric correction forcibly stretches the coordinates at anchor points, causing positional jumps or discontinuities in the velocity derivative of the calculated trajectory. This results in false high-frequency noise in the aligned signal, affecting the extraction of fault features. Therefore, this invention introduces a physical dynamics mechanism, modeling the coordinate system as an elastic slider driven by external forces. The instantaneous velocity of the train is considered the main driving force for the slider's movement, and the anchor point is considered the traction point of a virtual spring. When the current calculated position, obtained based on the integral of the train's instantaneous velocity, deviates from the absolute coordinates of the anchor point, the virtual spring generates flexible elastic force to smoothly correct the trajectory, thereby eliminating accumulated errors while ensuring the physical continuity of the trajectory.

[0039] Specifically, calculate the virtual correction acceleration at the current moment:

[0040]

[0041] In the formula, Indicates the first The virtual correction acceleration at each moment is used to eliminate the cumulative position error of the calculated trajectory relative to the real trajectory; Indicates the index of the current sampling time. Indicates the time step; This represents the system's inherent frequency parameter, which is used to determine the system's response speed to position error correction. Indicates the time elapsed since the current sampling time. The absolute coordinates of the nearest anchor point; Indicates the solution position at the previous time step; Indicates the time elapsed since the current sampling time. The trigger time of the most recent anchor point; This indicates the time window threshold for anchor point triggering; Represents the absolute value symbol. This represents the positional deviation between the absolute coordinates of the anchor point and the calculated position at the previous moment. The system is only subject to a positional deviation when the train is within a time window near the anchor point. The virtual correction acceleration is proportional to the deviation; the greater the deviation, the greater the virtual correction acceleration, thereby achieving rapid dynamic correction of accumulated errors.

[0042] It should be noted that the system's inherent frequency parameters This determines the system's response speed to correct anchor point position errors; based on the fact that vehicle depot entry inspections are typically slower than [the required speed], [the response speed is] lower. The characteristic is that the physical duration of the train passing through the effective induction zone of the magnet is usually within to Between these points, the response speed of the dynamic model must match the time scale: if The setting is too small, for example, below... The system's step response settling time will be much longer than the duration of the train passing the anchor point, causing the train to leave the anchor point before the virtual spring force has fully exerted its effect, resulting in insufficient correction; if The setting is too large, for example, exceeding... The system's frequency response bandwidth will cover the high-frequency noise band of the speed sensor, leading to the introduction of non-physical oscillation noise during low-speed, smooth motion. Therefore, in order to achieve fast and smooth convergence within a limited low-speed passage window, this invention sets an inherent frequency parameter. The range of experience is to In this embodiment, an empirical value that balances fast response and noise suppression is selected. Anchor point trigger time window threshold The effective time for the train wheels to pass through the magnetic induction zone is typically set. If the setting is too long, the train will still experience false traction acceleration after leaving the anchor point, compromising the accuracy of the trajectory. If the setting is too short, the correction process will be forcibly interrupted, resulting in residual errors. This embodiment combines the physical dimensions of the induction zone of a general axle-counting magnet with the average speed of the train passing through, and... Set as This ensures that the correction effect only takes effect at the moment of actual physical interaction.

[0043] For example, Figure 3 The diagram shows the virtual correction acceleration changing over time. It can be seen that the virtual correction acceleration only exhibits a pulse-like change near the anchor point triggering time, which verifies that the virtual spring only plays a flexible traction role within the time window with physical constraints, thereby effectively eliminating the cumulative drift caused by velocity integral.

[0044] Furthermore, the corrected velocity is updated based on the second-order dynamic equations:

[0045]

[0046] In the formula, Indicates the first Speed ​​after time correction, Indicates the index of the current sampling time; This indicates the corrected velocity from the previous moment; Indicates the first Virtual correction acceleration at any given moment; This represents the system's inherent frequency parameters; Indicates the train number Instantaneous velocity at a given moment; Indicates the time step; This represents the damping ratio parameter, used to determine the convergence smoothness of the correction process and suppress system oscillations. The rate of change of the corrected velocity depends on the resultant force of the virtual spring force and the virtual damping force, where the damping term... It exerted a soft constraint effect, forcing the corrected velocity As close as possible to the instantaneous speed of actual data collection. The changing trend is observed, while the dissipation characteristics of damping are used to smooth out high-frequency jitter noise in instantaneous velocity; and virtual correction acceleration... This serves as a correction term, intervening when the anchor point is triggered to correct the cumulative position error caused by the velocity integral. The two work together to ensure that the velocity curve is both smooth and continuous and accurate in position over long distances.

[0047] It should be noted that the damping ratio This determines the smoothness and convergence mode of the correction process. If the value is too small, for example less than 0.7, the system's response to errors, while quick, will produce non-physical oscillations, causing false positional fluctuations in the corrected trajectory near the anchor point; if... If the value is too large, such as greater than 1.5, the system's energy dissipates too quickly, leading to sluggish response and an inability to eliminate positional deviations within a finite time window, resulting in a tailing phenomenon. Therefore, to achieve the optimal balance between suppressing oscillations and achieving rapid convergence, a setting is necessary. The empirical range is 0.7 to 1.5. In this embodiment, to pursue the ideal trajectory characteristics with the fastest convergence and no overshoot, an empirical value of 1, i.e., the critical damping state, is selected to ensure that the corrected trajectory approximates the real physical position in a monotonically convergent manner, avoiding trajectory distortion caused by overshoot. In other embodiments, the implementer can set the value according to the distribution density of track anchor points and the noise level of the velocity sensor. Size: If the random noise of the speed sensor is large, the size can be appropriately increased. To enhance the filtering capability against high-frequency jitter; if the track anchor points are sparsely distributed, strict control is required. The value is close to 1 to prevent the positioning accuracy of subsequent intervals from being affected by excessively long oscillation convergence time.

[0048] Furthermore, update the corrected position based on the corrected velocity:

[0049]

[0050] In the formula, Indicates the first The corrected position at that moment. Indicates the index of the current sampling time; Indicates the first Speed ​​after time correction; Indicates the corrected position from the previous moment; Indicates the time step.

[0051] The corrected position at each moment constitutes a sequence of spatiotemporal mapping trajectories.

[0052] For example, Figure 4 This is a schematic diagram comparing the traditional rigid geometric correction trajectory, the spatiotemporal mapping trajectory sequence of this invention, and the actual physical trajectory. Figure 4 Traditional rigid geometric correction trajectories accumulate drift over time within the anchor point interval and exhibit significant vertical position jumps at the moment the anchor point is triggered. In contrast, the spatiotemporal mapping trajectory sequence of this invention smoothly converges to the real physical trajectory when passing through the anchor point due to the virtual correction acceleration. This verifies that this invention effectively avoids the position jump problem caused by traditional methods while eliminating accumulated errors.

[0053] The spatiotemporal resampling mapping module 300 is used to perform monotonicity processing on the spatiotemporal mapping trajectory sequence, and based on the monotonic spatiotemporal mapping trajectory sequence, to map the monitoring load data to spatial grid points in the train body coordinate system, generating spatially aligned monitoring data.

[0054] It should be noted that the original monitoring data is based on time-uniform sampling, while subsequent fault diagnosis requires fusion based on the physical parts of the train. Since the train speed fluctuates in real time, the physical distance swept by the train within the same time interval varies, causing nonlinear distortion in the spatial distribution of the data. This makes it impossible to directly align the data from different sensors. Furthermore, the spatiotemporal mapping trajectory sequence is located in the track coordinate system, while the fault characteristics are located in the train body coordinate system. The two are not directly equivalent. Therefore, this invention first performs monotonicity processing on the spatiotemporal mapping trajectory sequence, removes backtracking points, and constructs a single-investigation lookup table from position to time to ensure that each spatial position corresponds to a unique earliest arrival time. This allows the data from sensors scattered in different physical locations to be uniformly mounted onto the train body coordinate system, achieving strict alignment of physical positions.

[0055] Specifically, the spatial resampling resolution is set to A spatial grid point sequence is constructed with the train head as the origin. The relative coordinate values ​​of each spatial grid point in the spatial grid point sequence satisfy the expression: ,in , represents the index of a spatial grid point. Indicates the first The relative coordinates of each spatial grid point represent the position of each physical slice on the train body extending backward from the front, belonging to the relative coordinate system of the train body.

[0056] For trackside monitoring subsystems installed at different physical locations, calculate the target absolute coordinates of the track for each spatial grid point:

[0057]

[0058] In the formula, For the trackside monitoring subsystem The calculation of the first The target value of the absolute coordinates of the orbit corresponding to each spatial grid point; For the trackside monitoring subsystem Fixed installation coordinates on the track; Indicates the first The relative coordinates of each spatial grid point. When the train head travels onto the track... When in position, installed The sensors of the trackside monitoring subsystem are pointed at the car body, a distance from the front of the car. Sampling is performed at specific locations. This invention introduces fixed installation coordinates from a trackside monitoring subsystem. Successfully obtained the vehicle coordinates It was converted into the absolute orbital coordinates in the same domain as the spacetime mapping trajectory sequence. .

[0059] The spatiotemporal mapping trajectory sequence is processed for monotonicity, specifically by setting initial values ​​for the monotonic correction position. = Traverse the spatiotemporal mapping trajectory sequence. If the corrected position of the current moment in the spatiotemporal mapping trajectory sequence is... The monotonic correction position is smaller than the previous time step. This means that non-physical backoff noise has appeared, which forces the current monotonic correction position to be adjusted. Conversely, maintain It should be noted that this invention performs monotonicity processing on the spatiotemporal mapped trajectory sequence, filtering out all false reversing signals and ensuring that the monotonic processed spatiotemporal mapped trajectory sequence is strictly non-decreasing with time, thereby guaranteeing that the position-to-time mapping function has a unique inverse function.

[0060] Furthermore, for each target absolute track coordinate value, a reverse search and interpolation are performed on the spatiotemporal mapped trajectory sequence after monotonicity processing to find the precise time when the train arrives at the target absolute track coordinate value. Specifically: for each target absolute track coordinate value... Searching for the satisfying condition in the spatiotemporal mapping trajectory sequence after monotonicity processing index And use linear interpolation to calculate the target absolute coordinates of the train reaching the track. Precise time:

[0061]

[0062] In the formula, This indicates that the train has arrived for the trackside monitoring subsystem. The calculation of the first The target value of the absolute coordinates of the orbit corresponding to each spatial grid point The precise time; This indicates that the position interval inclusion condition is satisfied in the spatiotemporal mapped trajectory sequence after monotonicity processing. Sampling time index; This represents the first motif in the spatiotemporal mapped trajectory sequence after monotonicity processing. The monotonic correction position corresponding to each sampling time; This represents the first motif in the spatiotemporal mapped trajectory sequence after monotonicity processing. The monotonic correction position corresponding to each sampling time; For the trackside monitoring subsystem The calculation of the first The target value of the absolute coordinates of the orbit corresponding to each spatial grid point; Indicates the time step.

[0063] For each trackside monitoring subsystem, based on the precise time calculated for each trackside monitoring subsystem, the original monitoring data output by the trackside monitoring subsystem is resampled to obtain spatially aligned monitoring data that corresponds one-to-one with the spatial grid point sequence. This data reflects the physical state of the train body at each spatial grid point, eliminating the temporal distortion caused by fluctuations in train speed.

[0064] It should be noted that the spatial grid resolution determines the spatial accuracy of the aligned data. If the spatial grid resolution is set too high, minute fault features will be lost. In this embodiment, in order to meet the requirement of locating minute defects in the wheel tread, the spatial grid resolution is set to... In other embodiments, implementers can set the size of the spatial grid resolution according to the specific indicators of fault detection accuracy.

[0065] Figure 5 A waveform diagram of spatial alignment monitoring data, such as Figure 5 As shown, the infrared thermosensitive peaks and acoustic vibration peaks that were originally at different time points in the monitoring load data are now strictly aligned to the same spatial grid point in the generated spatially aligned monitoring data. This means that the data from different trackside monitoring subsystems have been accurately mapped to the same physical component of the train, verifying the ability of this invention to achieve spatial synchronization of multi-source heterogeneous data.

[0066] The alignment quality self-test output module 400 is used to calculate the alignment confidence residual during the train's passage based on the virtual correction acceleration and corrected speed at each moment. In response to the alignment confidence residual being less than or equal to a preset safety threshold, it outputs spatial alignment monitoring data.

[0067] It should be noted that since the correction force of the dynamic model indirectly reflects the quality of instantaneous velocity, if the speed measuring radar malfunctions or the wheels slip severely, the virtual spring will generate a huge correction force to counteract the erroneous instantaneous velocity. Although the calculated trajectory is geometrically pulled to the anchor point, its motion process seriously violates the laws of physics, making the spatiotemporal mapping relationship established thereby unreliable. Therefore, this invention utilizes the specific work done by the virtual correction acceleration to construct a self-checking index, evaluate the quality of this alignment, and prevent erroneously aligned data from flowing into downstream diagnostic systems and causing misjudgments.

[0068] Specifically, the alignment confidence residuals during this train passage process are calculated. :

[0069]

[0070] In the formula, This represents the alignment confidence residual during the train's passage, and its physical essence is the specific work done by the virtual correction acceleration, i.e., the work per unit mass. Indicates the first Virtual correction acceleration at any given moment; Indicates the first Speed ​​after time correction; Indicates the time step.

[0071] If the alignment confidence residual is less than or equal to a preset safety threshold, the alignment is determined to be successful. The spatial-temporal mapping trajectory sequence after monotonicity processing and the spatial alignment monitoring data of the trackside monitoring subsystems at different physical locations are packaged and output to the downstream fault diagnosis system. If the alignment confidence residual is greater than the preset safety threshold, the alignment is determined to be unsuccessful. An alarm signal is output and the unreliable monitoring data is discarded to block error propagation.

[0072] To ensure that the safety threshold can tolerate normal measurement noise while accurately intercepting abnormal data, this embodiment adopts a statistical calibration method based on historical data: a large amount of healthy historical data of known speed measuring devices working properly and trains running smoothly is collected, and the average value and standard deviation of the alignment confidence residuals during train passage in the healthy historical data are statistically analyzed; based on the statistical normal distribution law, the safety threshold is set as the average value plus 3 times the standard deviation. Once the real-time alignment confidence residual exceeds this limit, it means that a physical anomaly with a very low probability has occurred, such as wheel slippage or radar frame loss. Thus, the system can determine it as unreliable data and remove it, effectively ensuring the input quality of the downstream intelligent diagnostic system.

Claims

1. An intelligent and collaborative automated maintenance line system for rail transit vehicle depots, characterized in that: include: The multi-source data synchronous acquisition module is used to synchronously acquire motion drive data, constraint anchor point data and monitoring load data, and align all acquired data to a unified time base. The dynamic trajectory calculation module is used to input motion-driven data and constraint anchor point data into a pre-constructed dynamic model, and to perform flexible correction on the trajectory derived from the motion-driven data using the absolute coordinates of the anchor points in the constraint anchor point data. It calculates the virtual correction acceleration at the current moment, including: determining whether the absolute value of the time difference between the current sampling moment and the trigger moment of the nearest anchor point is less than the anchor point triggering time window threshold; in response to the absolute value of the time difference being less than the anchor point triggering time window threshold, determining the virtual correction acceleration at the current moment as the product of the square of the system's natural frequency parameter and the position deviation; otherwise, setting the virtual correction acceleration at the current moment to 0; wherein, the position deviation is the difference between the absolute coordinates of the nearest anchor point to the current sampling moment and the corrected position at the previous moment; The corrected velocity at the current moment is updated based on the virtual correction acceleration, and then the corrected position at the current moment is updated, forming a spatiotemporal mapping trajectory sequence with the corrected position at each moment. The spatiotemporal resampling mapping module is used to perform monotonicity processing on the spatiotemporal mapping trajectory sequence, and based on the monotonic spatiotemporal mapping trajectory sequence, it maps the monitoring load data to spatial grid points in the train body coordinate system to generate spatially aligned monitoring data. The alignment quality self-check output module is used to calculate the alignment confidence residual during the train's passage based on the virtual correction acceleration and corrected speed at each moment. In response to the alignment confidence residual being less than or equal to a preset safety threshold, it outputs spatial alignment monitoring data.

2. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 1, characterized in that, The motion drive data includes the instantaneous speed of the train; the constraint anchor point data includes the trigger time of the anchor point and the absolute coordinates of the anchor point; the monitoring load data includes the raw monitoring data output by the trackside monitoring subsystems installed at different physical locations.

3. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 1 or 2, characterized in that, The method for obtaining the anchor points is as follows: Axle-counting magnets installed at fixed intervals along the track monitor the pulse signals as the wheels pass by. When the rising edge of the pulse is detected, the trigger time is recorded, and the absolute coordinates are obtained according to the preset physical installation position of the magnets. The trigger time and the absolute coordinates are combined to define the anchor point.

4. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 1, characterized in that, The step of updating the corrected velocity at the current moment based on the virtual correction acceleration includes: Calculate the difference between the corrected speed at the previous moment and the instantaneous speed of the train at the current moment, as the speed error term; calculate twice the product of the system's natural frequency parameter, damping ratio, and the speed error term, as the damping adjustment term; calculate the difference between the virtual corrected acceleration at the current moment and the damping adjustment term, multiply the result by the time step to obtain the speed correction amount; add the corrected speed at the previous moment to the speed correction amount to obtain the corrected speed at the current moment.

5. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 1, characterized in that, The update of the corrected position at the current moment includes: Calculate the product of the corrected velocity and the time step at the current moment as the position increment; add the corrected position at the previous moment to the position increment to obtain the corrected position at the current moment.

6. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 1, characterized in that, The monotonic processing of the spatiotemporal mapping trajectory sequence includes: Set an initial value for the monotonic correction position, traverse the spatiotemporal mapping trajectory sequence, and in response to the current corrected position being less than the previous monotonic correction position, set the current monotonic correction position to the previous monotonic correction position; in response to the current corrected position being greater than or equal to the previous monotonic correction position, set the current monotonic correction position to the current corrected position.

7. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 1, characterized in that, The process of mapping the monitoring load data to spatial grid points in the train body coordinate system includes: A spatial grid point sequence with the train head as the origin is constructed. For each trackside monitoring subsystem that outputs monitoring load data, the fixed installation coordinates of the trackside monitoring subsystem are added to the relative coordinate values ​​of each spatial grid point in the spatial grid point sequence to obtain the target absolute track coordinates for each spatial grid point of the trackside monitoring subsystem.

8. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 7, characterized in that, The generated spatial alignment monitoring data includes: For each target value of absolute track coordinates, a search is performed in the spatiotemporal mapping trajectory sequence after monotonicity processing to determine the sampling time index that makes the monotonic correction position fall within the position interval containing the target value of absolute track coordinates. Based on the sampling time index, the precise time when the train arrives at the target value of absolute track coordinates is calculated using linear interpolation. Based on the precise time, the monitoring load data is resampled to obtain spatially aligned monitoring data that corresponds one-to-one with the spatial grid point sequence.

9. The intelligent collaborative automated maintenance line system for rail transit vehicle depots according to claim 1, characterized in that, The calculation of the alignment confidence residual during the train's passage based on the virtual corrected acceleration and corrected velocity at each moment includes: Calculate the absolute value of the product of the virtual corrected acceleration, the corrected velocity, and the time step at each moment, and use it as the virtual work done at that moment; sum up the virtual work done at all moments during the train's passage to obtain the alignment confidence residual for this train's passage.

Images