On-line detection system and method for gear of rack rail train

Abstract

The invention relates to the technical field of state monitoring of rail transit equipment, and discloses a rack rail train gear online detection system and method, and the system comprises a laser sensor array module; a visual perception assembly module; a trigger module; and a data processing module. The method comprises the following steps: synchronously acquiring vertical distance data and an image sequence; applying an optical flow algorithm to track an image sequence to calculate a gear rotation angle; installing errors and vehicle vibration are separated through cross-correlation and low-pass filtering, and distance data are corrected; reconstructing the full-circle geometric contour of the gear according to the rotation angle; and finally, geometric deviation and surface damage are judged based on contour and image analysis. According to the invention, through multi-modal data fusion and error decoupling, high-precision automatic on-line detection of the gear of the rack rail railway vehicle under the dynamic passing working condition of the vehicle is realized, and the problems that the traditional off-line dismounting detection efficiency is low and early damage cannot be found in time are solved.

Description

Technical Field

[0001] This invention relates to the field of rail transit equipment condition monitoring technology, specifically to an online detection system and method for gears in rack trains. Background Technology

[0002] Gear-rack railway vehicles are special rail transport vehicles that rely on the meshing of gears and racks for transmission, enabling them to operate on steep slopes. As a core component ensuring the safe and stable operation of the vehicle, the transmission gears, under long-term heavy-load service, are susceptible to problems such as tooth surface damage, tooth root cracks, or out-of-roundness deviations due to fatigue wear and other factors. Therefore, it is essential to conduct regular and precise geometric dimensional and surface integrity inspections of the transmission gears to promptly identify potential problems and develop maintenance plans.

[0003] Currently, the industry generally adopts offline testing methods for the inspection of gears in rack railway vehicles. This method requires the vehicle to be stopped and placed in a maintenance depot, where maintenance personnel disassemble the gearbox from the bogie and then send the removed gears to a coordinate measuring machine or gear measurement center for measurement. Although the offline method can achieve high inspection accuracy, its process is lengthy, requires a large amount of manpower and time, and results in prolonged vehicle downtime, significantly reducing vehicle utilization efficiency and making it difficult to adapt to the high-density, routine maintenance rhythm of modern rail transit.

[0004] The fundamental limitation of existing inspection methods lies in their offline and static operation mode. Because gears must be disassembled, the inspection cycle is long, making it difficult to achieve high-frequency monitoring of gear conditions and hindering the timely detection of early, minor damage. Furthermore, directly applying inspection equipment to online inspection on the production line also faces technical challenges. For example, when a vehicle dynamically passes by, the vehicle's own vibrations mix with the actual contour signal of the gears, creating significant measurement interference. To accurately reconstruct the three-dimensional contour of the gears, their precise rotation angles must be acquired simultaneously, but achieving reliable non-contact dynamic angle measurement without additional encoders is extremely difficult. Moreover, when multiple sensors are used for collaborative measurement, even slight differences in their installation positions can introduce persistent system errors. These factors all hinder the engineering implementation of high-precision online inspection solutions.

[0005] Therefore, this invention proposes an online detection system and method for gears in geared trains to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an online detection system and method for gears in rack trains, solving the technical challenge of how to achieve non-contact, high-precision, and automated online detection of gear geometry and surface damage through multimodal data fusion and error decoupling under dynamic vehicle passage conditions.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first aspect of the present invention provides an online detection system for gears in a geared train, comprising: A laser sensor array module includes at least two laser displacement sensors linearly distributed along the circumference of the gear rotation. The laser displacement sensors are used to continuously acquire vertical distance data of the gear tooth surface relative to the ground. The visual perception component module includes a structured light projector and an industrial camera. The structured light projector is used to project an artificial texture pattern onto the side of a gear, and the industrial camera is used to synchronously acquire a sequence of image data of the side of the gear containing the artificial texture pattern. The trigger module is used to monitor the position status of the vehicle wheelsets and send start and stop signals; The data processing module is communicatively connected to the laser sensor array module, the visual perception component module, and the trigger module. It is used to respond to the start signal to control the laser sensor array module and the visual perception component module to synchronously acquire and process the vertical distance data and the gear side image data sequence to obtain the gear's geometric contour and surface damage features.

[0008] Preferably, in the laser sensor array module, the detection ranges of two adjacent laser displacement sensors in the circumferential direction of gear rotation physically overlap to form an overlapping field of view.

[0009] Preferably, the data processing module is further configured to: Select distance data segments from adjacent laser displacement sensors within the overlapping field of view; Perform waveform cross-correlation analysis on the distance data segment to separate the DC deviation caused by sensor installation error; The vertical distance data is corrected using the DC deviation.

[0010] Preferably, the data processing module is further configured to: An optical flow algorithm is applied to track the displacement features of the artificial texture pattern in the gear side image data sequence in order to calculate the global instantaneous angular velocity of the gear. Numerical integration is performed on the global instantaneous angular velocity to generate a sequence of gear rotation angles.

[0011] Preferably, the data processing module is further used for: Establish a mapping relationship between the vertical distance data and the gear rotation angle sequence, and convert the vertical distance data in the time domain into angle domain data.

[0012] Preferably, the data processing module is further used for: A digital low-pass filter is applied to the continuous distance data after the DC deviation has been eliminated to extract the AC vibration component caused by the vehicle suspension system and construct a vehicle vertical vibration data model. The vertical distance data is further corrected using the vehicle vertical vibration data output from the vehicle vertical vibration data model to obtain the corrected vertical distance data.

[0013] Preferably, the data processing module is further used for: Based on the gear rotation angle sequence, the corrected vertical distance data is mapped to a two-dimensional polar coordinate system to reconstruct the full circumference geometric profile of the gear.

[0014] Preferably, the data processing module is further used for: Based on the full circumferential geometric profile of the gear, the maximum radial runout value is calculated, and key geometric parameters, including tooth thickness and addendum circle diameter, are extracted. The maximum radial runout value is compared with the key geometric parameters and a preset reference parameter to obtain the geometric parameter deviation.

[0015] Preferably, the data processing module is further used for: Identify regions in the full circumferential geometry of the gear that exhibit abrupt changes; Based on the angular index of the abnormal mutation region, the image data at the corresponding time is retrieved and speckle removal and edge detection are performed to identify surface damage features, including tooth surface wear and tooth root cracks.

[0016] A second aspect of the present invention provides an online detection method for gears in a rack train, comprising the following steps: In response to the start signal sent by the trigger module, the vertical distance data of the gear tooth surface relative to the ground and the image data sequence of the gear side surface containing artificial texture patterns are collected synchronously. An optical flow algorithm is applied to track the gear side image data sequence, and the gear rotation angle sequence is calculated and integrated to generate the gear rotation angle sequence. Distance data segments from adjacent laser displacement sensors within the overlapping field of view are selected. DC deviation is separated through waveform cross-correlation analysis. The data with eliminated DC deviation is then low-pass filtered to extract AC vibration components. The DC deviation and AC vibration components are used to correct the vertical distance data to obtain corrected vertical distance data. Based on the gear rotation angle sequence, the corrected vertical distance data is mapped to a two-dimensional polar coordinate system to reconstruct the full circumference geometric profile of the gear; Based on the full circumference geometric contour of the gear, the geometric parameter deviation is calculated, and the surface damage characteristics are identified by analyzing the images corresponding to the abnormal change areas. Finally, the health status and damage level of the gear are comprehensively determined.

[0017] This invention provides an online detection system and method for gears in geared trains. It has the following beneficial effects: 1. This invention achieves seamless splicing measurement of the entire circumference of the gear by setting up a laser sensor array module linearly distributed along the circumference of the gear's rotation and making the detection ranges of adjacent sensors overlap. Furthermore, by performing waveform cross-correlation analysis on two sets of independently acquired vertical distance data within the overlapping area, this invention can accurately calibrate and separate the systematic DC deviation introduced by inconsistent sensor installation positions. This data-driven self-calibration mechanism effectively overcomes the interference of on-site installation accuracy on the measurement results, avoids cumbersome physical adjustments, and thus obtains high-precision original data of the gear's geometric profile under a unified benchmark without interrupting vehicle operation, providing a reliable data foundation for subsequent accurate analysis of geometric parameters such as out-of-roundness.

[0018] 2. This invention creates stable, high-contrast visual tracking features by introducing a structured light projector onto the side of the gear using a visual perception component module, and simultaneously acquiring image sequences using an industrial camera. This design cleverly solves the problem of feature loss and difficulty in motion estimation caused by oil stains or reflections on the side of the gear in traditional methods. The data processing module uses an optical flow algorithm to track the displacement of the artificial texture, accurately calculating the global instantaneous angular velocity of the gear during the detection process, and integrating to generate a high-resolution rotation angle sequence. This non-contact angle measurement method eliminates the dependence on installing a rotary encoder, greatly simplifies system deployment, achieves accurate dynamic capture of any gear rotation state, and ensures precise mapping between distance data and angular coordinates.

[0019] 3. This invention achieves spatiotemporal alignment between the calculated gear rotation angle sequence and the collected vertical distance data, and decouples and separates the high-frequency tooth profile undulation signal and low-frequency vibration interference signal in the full-circumference contour data. Specifically, by performing digital low-pass filtering on the data after eliminating installation errors, the low-frequency AC vibration component generated by the vehicle suspension system during dynamic driving can be effectively extracted. Using this vibration component to dynamically compensate and correct the measurement data, this invention successfully eliminates the influence of vehicle bumps on contour measurement, restoring the true geometric contour of the gear relative to its own center of rotation. This multimodal data fusion and error separation strategy ensures reliable identification and assessment of minute damage features such as tooth surface wear and tooth root cracks, even under complex conditions of dynamic vehicle passage. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of an online detection system for gears in a geared train according to the present invention. Figure 2This is a flowchart of an online detection method for gears in a rack-and-rail railway vehicle according to the present invention; Figure 3 This is a schematic diagram of the multimodal data fusion and error separation process of the present invention.

[0021] Among them, 100 is a laser sensor array module; 101 is a laser displacement sensor; 200 is a visual perception component module; 201 is an industrial camera; 202 is a structured light projector; 300 is a trigger module; and 400 is a data processing module. Detailed Implementation

[0022] The technical solutions in 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 embodiments of the present invention, and not all embodiments. 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.

[0023] See attached document Figure 1 This invention provides an online detection system for gears in a rack train, comprising: a laser sensor array module 100, a vision perception component module 200, a trigger module 300, and a data processing module 400. These modules are connected via an industrial fieldbus or high-speed Ethernet to form a measurement system with synchronous data acquisition, real-time processing, and closed-loop feedback capabilities.

[0024] The laser sensor array module 100 is installed below the track of the vehicle entry section, located within a preset detection area. The laser sensor array 100 includes at least two laser displacement sensors 101 linearly distributed along the circumference of the gear rotation. The emitting end of each laser displacement sensor 101 is vertically upward, and its measuring beam covers the motion trajectory of the gear tooth tip and tooth valley, used to continuously collect the vertical distance data of the gear tooth surface relative to the ground.

[0025] The visual perception module 200 is mounted on the upper side of the detection area via a rigid bracket, and its field of view covers the entire circumference of the gear's side surface. The visual perception module 200 includes an industrial camera 201 and a structured light projector 202. The structured light projector 202 is configured to project an artificial textured pattern with high contrast characteristics onto the gear's side surface. The industrial camera 201 is configured to operate in a high frame rate mode, simultaneously acquiring a sequence of image data of the gear's side surface containing the aforementioned artificial textured pattern.

[0026] The trigger module 300 is located at the entrance of the detection area and typically employs an infrared through-beam or laser interruption sensor. The trigger module 300 monitors the position of the vehicle's wheelsets and sends corresponding start and stop signals to the data processing module 400 when the vehicle enters and leaves the detection area, thereby automating the detection process.

[0027] The data processing module 400, as the core computing and control hub of the system, is communicatively connected to the laser sensor array module 100, the vision perception component module 200, and the trigger module 300. The data processing module 400 integrates high-precision clock synchronization logic to uniformly control the acquisition timing of the laser displacement sensor 101 and the industrial camera 201, ensuring that distance data and image data have aligned timestamps.

[0028] See attached document Figure 2 This invention provides an online inspection method for gears in rack-rail railway vehicles, based on the aforementioned online inspection system for rack-rail train gears, comprising the following steps: S10, System Initialization and Self-Test. Before the vehicle enters the warehouse, the data processing module 400 performs status checks and parameter calibrations on each system component. The laser sensor array module 100 performs zero-point calibration, confirming the initial distance reading with the ground as a reference plane. The vision perception component module 200 adjusts the exposure parameters of the industrial camera 201 and illuminates the structured light projector 202 to confirm that the artificial texture pattern is projected normally. The data processing module 400 checks the communication link status and storage space.

[0029] S20, Trigger Start-up and Projection Enhancement. When a rack railway vehicle enters the depot section, and its wheels pass through the detection area entrance, the trigger module 300 senses the wheelset position and sends a trigger signal to the data processing module 400. The data processing module 400 responds to the trigger signal, controlling the structured light projector 202 to continuously operate, forming a high-contrast artificial texture pattern on the side of the gear that rigidly rotates with the gear. Simultaneously, it sends synchronous acquisition commands to the laser sensor array module 100 and the visual perception component module 200.

[0030] S30, Multimodal Data Synchronous Acquisition. The vehicle travels along the entry track, and the wheels drive the gears to rotate synchronously and passively via the drive shaft. During this process, each laser displacement sensor 101 in the laser sensor array module 100 continuously acquires vertical distance data of the gear tooth surface relative to the ground at a high sampling rate; the industrial camera 201 synchronously and continuously acquires image data of the gear side containing artificial texture patterns. The data processing module 400 uses a unified hardware clock signal to stamp all acquired distance data and image data with high precision and stores them in time sequence to the local high-speed buffer storage area.

[0031] S40, Data Processing and Contour Reconstruction. The data processing module 400 performs hierarchical analysis and fusion on the acquired raw data. First, it uses an optical flow algorithm to track the displacement characteristics of the artificial texture pattern in the image data, calculates the global instantaneous angular velocity of the gear, and integrates to generate a gear rotation angle sequence corresponding to the timestamp. Second, it selects vertical distance data segments from adjacent laser displacement sensors 101 within the overlapping field of view, and separates the DC deviation caused by sensor installation errors and the AC vibration components caused by the vehicle suspension system through waveform cross-correlation analysis, constructing a vehicle vertical vibration data model. Finally, it uses the vehicle vertical vibration data to correct the amplitude of the raw distance data, and maps the corrected vertical distance data to a polar coordinate system based on the gear rotation angle sequence, reconstructing the full-circumference geometric contour of the gear after eliminating environmental interference.

[0032] S50, Comprehensive Analysis and Damage Assessment. Based on the reconstructed full-circumference geometric profile of the gear, the data processing module 400 calculates the maximum radial runout value of the gear to determine the out-of-roundness index, and at least includes extracting key geometric parameters such as tooth thickness and addendum circle diameter and comparing them with standard benchmarks. Simultaneously, for regions exhibiting abnormal abrupt changes in the geometric profile, the data processing module 400 retrieves the corresponding image data according to the angle index, performs speckle removal and edge detection on the images, and identifies surface damage features including at least tooth surface wear and tooth root cracks. Combining the geometric parameter deviations and surface damage identification results, the health status and damage level of the gear are determined.

[0033] S60, Report Generation and Data Archiving. After the inspection process is completed, the data processing module 400 automatically generates a standardized inspection report. The report includes the vehicle number, inspection time, out-of-roundness value, abnormal geometric parameters, and high-resolution images of the damaged areas. The inspection report is uploaded to the vehicle operation and maintenance management system in real time via the industrial fieldbus and displayed on the on-site display terminal. The raw collected data and the processed structured data are compressed and archived to the local storage unit for subsequent historical traceability and trend analysis.

[0034] To further clarify the technical details of the system and method in this embodiment, the specific implementation methods, principles and technical contents of each module and method will be described in detail below.

[0035] Before the system is put into operation, the data processing module 400 will execute a series of initialization and self-test procedures, including zero-point calibration of the laser sensor array module 100, adaptive adjustment of the exposure parameters of the vision perception component module 200, and loading the standard geometric parameters of the vehicle under test.

[0036] See attached document Figure 1 and Figure 2In order to achieve high-precision online gear detection under passive vehicle driving conditions, the laser sensor array module 100 and the vision perception component module 200 adopt specific photoelectric parameter configuration and physical space topology design.

[0037] The laser sensor array module 100 is configured to include 2 to 3 laser displacement sensors 101. These sensors are linearly arranged along the track extension direction and fixed in pre-set mounting slots on both sides of the track at the entrance of the depot section. The optical axis of the emitting end of each laser displacement sensor 101 is strictly perpendicular to the horizontal ground and set upwards, so that when a vehicle passes by, its measuring beam covers the motion trajectory of the gear tooth tip and tooth valley. This motion trajectory refers to the vertical projection path of the gear outline as the gear rolls with the wheel, covering the entire geometric feature area from the root circle to the tip circle. During the detection process, when the vehicle travels along the entrance track, the wheel drives the gear to rotate synchronously and passively through the drive shaft. Based on this, each laser displacement sensor 101 continuously collects the vertical distance data of the gear tooth surface relative to the ground. This vertical distance data is a set of one-dimensional scalar sequences that vary with time. Its physical meaning represents the absolute physical height from the laser emitting end reference plane to the instantaneous surface point of the gear, directly reflecting the cross-sectional profile undulation of the gear in the unfolding direction.

[0038] To meet the measurement requirements of gear micron-level morphology, the laser displacement sensor 101 is a high-precision laser triangular reflector sensor with a measurement accuracy set to no higher than 0.01 mm. Its effective measurement range covers 50 mm to 500 mm to accommodate variations in ground clearance for gears of different diameters. The light source wavelength of each laser displacement sensor 101 is selected as 650 nm visible red light, which facilitates spot alignment during installation and debugging while avoiding common infrared thermal radiation interference from the vehicle's underside. Its sampling frequency is set to no less than 100 Hz, preferably 2 kHz, to ensure a sufficient density of sampling points during dynamic vehicle operation.

[0039] In terms of spatial layout, two adjacent laser displacement sensors 101 (logically labeled as...) and There is physical overlap in the detection range along the circumferential direction of gear rotation. Let the effective measurement arc length of a single laser displacement sensor 101 on the gear's rotation trajectory be... The physical mounting distance in the horizontal direction between the centers of the optical axes of two adjacent sensors is The arc length of the overlapping field of view between the two is... Determined by the following geometric relationships: ; This embodiment controls the installation spacing. This results in a higher overlap rate in the detection ranges of adjacent sensors. No less than 30%. Overlap rate. The definition of is: ; Set overlap rate The purpose is twofold: firstly, to ensure that the cross-sectional data of the entire circumference is covered without any blind spots during the rotation of the gear, thus achieving a complete record of the gear's morphology; secondly, within the overlapping field of view, the same tooth surface feature will pass through two sensors successively, which provides the necessary redundant data foundation for subsequently separating the sensor installation height error and vehicle vibration displacement by comparing waveform features.

[0040] The visual perception component module 200 is rigidly mounted on a column above the detection area. To construct a clear 3D visual measurement benchmark, the structured light projector 202 and industrial camera 201 in the visual perception component module 200 employ an anti-interference design combining spectral separation and spatial avoidance. The structured light projector 202 is configured to emit near-infrared light with a center wavelength of 850nm, which differs from the 650nm visible light emitted by the laser displacement sensor 101. The structured light projector 202 integrates an optical shaping element configured to project a high-contrast artificial texture pattern onto the side of the gear. This artificial texture pattern is typically a random speckle or non-periodic stripe, its function being to construct optical feature points that can be tracked by image algorithms on a smooth metal surface lacking natural texture. For the specific drive circuit design of the structured light projector 202, those skilled in the art can use a constant current source driver board combined with PWM dimming control, which is well-known in the field and will not be elaborated further here.

[0041] The industrial camera 201 uses a high-definition industrial CCD or CMOS camera with a resolution of 20 megapixels or higher, equipped with a fixed-focus lens with a focal length of 25mm to 50mm, and features a global shutter function to eliminate motion blur. The camera lens is positioned to capture the gear's rotation trajectory, forming an angle of 30° to 60° with the vertical emission direction of the laser displacement sensor 101. This angle setting not only prevents the direct beam from the laser displacement sensor 101 from entering the camera's field of view but also ensures that the camera can capture the complete details of the gear's tooth surface and root from a side-view perspective. Furthermore, a narrow-band filter for the 850nm wavelength is mounted in front of the lens of the industrial camera 201 to physically block ambient stray light and interference from the 650nm red light of the laser displacement sensor 101. With this configuration, the industrial camera 201 simultaneously acquires a sequence of gear side image data containing the aforementioned artificial textured pattern. The gear side image data sequence is a collection of multiple consecutively captured two-dimensional digital images. Each frame clearly records the instantaneous state of the gear side covered by the projected artificial texture pattern, which is used for subsequent analysis of the gear's rotation angle and phase information. The acquisition frame rate of the industrial camera 201 is set to no less than 30fps, preferably above 100fps, to adapt to the optical flow algorithm's requirements for tracking minute inter-frame displacements.

[0042] The trigger module 300 uses an industrial-grade infrared through-beam or laser interruption sensor (such as an infrared through-beam switch) and is installed at the entrance of the detection area. The signal output terminal of this module, the external trigger terminal of the laser displacement sensor 101, and the hard trigger input terminal of the industrial camera 201 are all connected to the I / O control interface of the data processing module 400, forming a hardware synchronous control link.

[0043] In the actual workflow, the trigger module 300 is used to monitor the position status of the vehicle wheelsets in real time. When a rack railway vehicle enters the depot section and its wheels block the trigger module 300, the module senses the wheelset position and sends a start signal to the data processing module 400. The data processing module 400 responds to this hardware trigger signal by immediately generating a unified synchronization pulse, which is then distributed to: The structured light projector 202 is controlled to work continuously to form a high-contrast artificial texture pattern on the side of the gear that rotates rigidly with the gear. The laser sensor array module 100 drives each laser displacement sensor 101 to start high-speed ranging. The visual perception component module 200 drives the industrial camera 201 to start synchronous exposure and image acquisition.

[0044] In this way, the system ensures that the multimodal data stream (vertical distance data and gear side image data sequence) has a strictly aligned time reference from the first frame / first point acquired. When the vehicle has completely passed through and left the detection area, the trigger module 300 sends another stop signal to the data processing module 400 to realize the automatic start and stop control of the entire detection process.

[0045] See attached document Figures 1-3 The data processing module 400 receives the gear side image data sequence synchronously acquired by the industrial camera 201, and performs angle calculation based on it.

[0046] The data processing module 400 preprocesses and extracts the region of interest for each frame of the gear side image data sequence. To reduce computational load and remove background interference, the data processing module 400 uses pre-calibrated gear shaft center coordinates. The structured light projection range is defined by extracting a ring-shaped region in the image containing only artificial texture patterns as the computational domain.

[0047] For pixel data within the region of interest, the data processing module 400 applies an optical flow algorithm to track the displacement features of the artificial texture pattern in the gear side image data sequence. The optical flow algorithm is a computer vision technique that infers the instantaneous velocity vector field of pixels on a two-dimensional image plane by analyzing the grayscale change patterns of pixels between two adjacent frames. This algorithm is based on the assumption of constant brightness, meaning that the same point on the artificial texture pattern, which serves as a displacement feature, will exhibit constant brightness within extremely short sampling intervals. The brightness value within remains constant. Assume... For a moment Pixel coordinates If the brightness value at a given location is given, then the optical flow constraint equation can be expressed as: ; in, , and These represent the spatial gradient and temporal gradient of the image in the horizontal and vertical directions, respectively. and These represent the instantaneous velocity components of a pixel in the horizontal and vertical directions of the image plane, respectively. The data processing module 400 obtains the velocity vector of each pixel within the region of interest by solving this equation within a local window. This velocity vector is the representation of the displacement characteristics of the corresponding point on the artificial texture pattern on the image plane.

[0048] The data processing module 400 converts the translational velocity vector on the image plane into the rotational angular velocity of the gear about its axis. For any valid pixel within the region of interest... Its relative to the center of rotation instantaneous angular velocity The calculation is as follows: ; To eliminate local measurement noise, the data processing module 400 statistically averages the angular velocities of all valid pixels within the region to obtain the current time. Global instantaneous angular velocity of the gear .

[0049] After obtaining continuous angular velocity values, the data processing module 400 processes the global instantaneous angular velocity. Perform numerical integration to generate a sequence of gear rotation angles. Here, integration refers to the mathematical process of converting instantaneous velocity quantities into cumulative position quantities by accumulating angle increments within a time step. Set the initial time. Angle Then at time Cumulative rotation angle The calculation formula is as follows: ; Through the above recursive calculation, the system outputs a set of gear rotation angle sequences that correspond one-to-one with the image timestamps. This gear rotation angle sequence precisely describes the cumulative rotation angle of the gear from the moment it enters the detection zone.

[0050] To achieve spatial alignment of multimodal data, the data processing module 400 establishes a mapping relationship between the vertical distance data collected by the laser displacement sensor 101 and the gear rotation angle sequence. Since the laser sampling frequency is higher than the image frame rate, the data processing module 400 adjusts the laser displacement sensor 101 at different times... Collected vertical distance data Angle matching is achieved using linear interpolation. ; in, Represents the calculated time The corresponding gear rotation angle interpolation; and These represent the nth and nth elements in the image data sequence, respectively. Frame and the The cumulative rotation angle of the gear calculated at each frame time; and These represent the first and second images in the image data sequence, respectively. Frame and the The frame's acquisition timestamp; This indicates the specific time when the laser displacement sensor acquires the vertical distance data, and satisfies... This step converts the vertical distance data in the time domain into angular domain data, ensuring that each laser measurement point has accurate angular coordinates.

[0051] See attached document Figures 1-3 After acquiring measurement data aligned with the angle sequence, the data processing module 400 performs in-depth cleaning and reconstruction of the original distance data. The original distance data refers to the data directly acquired by the laser displacement sensor 101, before systematic errors and dynamic interference have been eliminated, and is expressed in angles... The set of vertical distance data values ​​indexed, the original distance data is denoted as .

[0052] The data processing module 400 selects a data segment of the vertical distance between adjacent laser displacement sensors 101 within the overlapping field of view. This data segment refers to the data segment between two physically adjacent sensors within the same gear rotation angle range. The two independent measurement sequences collected separately are denoted as follows: and .

[0053] For the two extracted data segments, the data processing module 400 performs waveform cross-correlation analysis. Waveform cross-correlation is used to measure the similarity and phase shift between two signal waveforms. Cross-correlation function. The definition is as follows: ; in, This is the angular phase lag. It is solved... The maximum value is used by the data processing module to verify the accuracy of angle synchronization at 400°. If the maximum value corresponds to... If the value deviates significantly from zero, the system will correct the phase of the angle sequence to ensure that the two sets of data segments are strictly aligned in time.

[0054] After phase alignment is completed, the data processing module 400 further separates the error term using the results of cross-correlation analysis. The data processing module 400 calculates the amplitude difference between the two sets of waveforms within the overlapping region, separating the DC deviation caused by sensor installation errors. The DC deviation is a constant measurement difference caused by the inconsistent physical installation heights of adjacent sensors. Its calculation is based on the average amplitude difference between two sets of aligned waveforms: ; in, This represents the number of sampling points within the overlapping region. The data processing module 400 utilizes the calculated... The measurement references of each sensor are uniformly shifted, and the data segments are stitched together to generate a continuous, unified reference data sequence covering the entire circumference, denoted as... .

[0055] Based on this, the data processing module 400 separates the AC vibration component caused by the vehicle's suspension system. The AC vibration component refers to the low-frequency displacement fluctuations in the vertical direction caused by the elastic action of the vehicle's suspension system and track irregularities during vehicle operation. The data processing module 400 extracts the benchmark unified data sequence. By analyzing the low-frequency trend term in the data and combining it with the frequency domain characteristics of the full-cycle data, a vehicle vertical vibration data model is constructed.

[0056] The process of constructing the vehicle vertical vibration data model is as follows: The data processing module 400 uses spectrum analysis technology to identify signal characteristics. Since the tooth profile undulations of gears are high-frequency signals (their fundamental frequency depends on the number of teeth and rotational speed), while the vehicle body float caused by the vehicle suspension system is a low-frequency signal (typically low frequency), the process is as follows: The two are significantly separable in the frequency domain. The data processing module 400 is designed with a cutoff frequency. A digital low-pass filter located between the vehicle body vibration frequency and the gear meshing frequency. The data processing module 400 will unify the baseline data sequence. The data is transformed to the frequency domain, multiplied by the frequency response function of the filter, and then inversely transformed back to the time domain to extract the continuous low-frequency fluctuation curve. This low-frequency fluctuation curve is the vehicle's vertical vibration data model, and its mathematical expression is as follows: ; in, This refers to the output vehicle vertical vibration data; Indicates Fourier transform, This represents continuous distance data after DC bias has been eliminated. The vehicle vertical vibration data output by this vehicle vertical vibration data model is the vehicle vertical vibration data. It is a continuous function that varies with angle, accurately describing the dynamic floating trajectory of the vehicle wheelset center during the detection process.

[0057] The data processing module 400 utilizes vehicle vertical vibration data ( and the aforementioned DC deviation For the raw distance data Amplitude correction is performed to generate corrected vertical distance data. This refers to a clean data sequence that has been freed from sensor position errors and vehicle dynamic vibration interference, reflecting only the geometrical distance change of the gear tooth surface relative to the gear rotation center. The correction calculation is as follows: ; The corrected vertical distance data It purely reflects the change in geometric distance between the gear tooth surface and the gear rotation center, eliminating the influence of foundation settlement error and vehicle dynamic bumps, and providing high signal-to-noise ratio contour data support for subsequent polar coordinate mapping and out-of-roundness determination.

[0058] See attached document Figures 1-3 After error separation and data correction are completed, the data processing module 400 performs comprehensive analysis and damage assessment.

[0059] The data processing module 400 first generates the gear rotation angle sequence based on the aforementioned steps. The one-dimensional corrected vertical distance data Mapping to a two-dimensional polar coordinate system to reconstruct the full circumferential geometry of the gear. Let the measurement reference radius be... Then the first outline Polarity of each sampling point The calculation formula is: ; The resulting point set This forms a closed circumferential geometric profile of the gear, which objectively reflects the true physical cross-sectional shape of the gear after eliminating sensor installation errors and vehicle dynamic vibrations.

[0060] Based on the reconstructed full-circumference geometric profile of the gear, the data processing module 400 calculates the maximum radial runout value of the gear to determine the out-of-roundness index. The data processing module 400 iterates through the extreme diameter sequence of the full-circumference profile data to obtain the maximum extreme diameter. With minimum polarity The difference between the two is defined as the maximum radial runout value. : ; The calculations obtained here Numerical quantification measures the degree to which the overall shape of the gear deviates from the ideal circumference, and is used to characterize whether there is eccentric or elliptical deformation.

[0061] Simultaneously, the data processing module 400 extracts key geometric parameters, including at least tooth thickness and addendum circle diameter, from the gear's full circumferential geometric profile. For the addendum circle diameter, the data processing module 400 identifies the set of local maxima corresponding to each tooth tip position in the polar diameter sequence, calculates the arithmetic mean of this set, and multiplies it by 2. For the tooth thickness calculation, the data processing module 400 sets the standard pitch circle radius. Find the angular coordinates of the intersection points of the profile curve and the pitch circle. Let the angles of the intersection points of the tooth profiles on both sides of a single tooth and the pitch circle be... and Then the tooth thickness at that location The calculation is as follows: ; The data processing module 400 compares the calculated values ​​of the addendum circle diameter and tooth thickness with the standard reference parameters of that gear model stored in the database. Here, the database refers to a relational database (such as SQLite or MySQL) deployed in the local storage unit, which pre-inputs and stores theoretical design parameters (including standard module, standard tooth thickness, tolerance range, etc.) and historical maintenance standards for various locomotive gear models. The difference between the comparisons (i.e., the measured value minus the standard value) constitutes the geometric parameter deviation. Geometric parameter deviation is quantitative data reflecting gear wear (such as tooth surface thinning) or dimensional deviations.

[0062] After completing the geometric dimension analysis, the data processing module 400 performs differential analysis on the full circumference geometric profile of the gear. When the rate of change (derivative) of the extreme diameter in a certain region of the profile exceeds a preset threshold for the normal tooth profile gradient, the data processing module 400 marks it as a region exhibiting an abnormal abrupt change. For each region of abnormal abrupt change, the data processing module 400 uses its corresponding angle index... The system retrieves the gear side image data acquired at the corresponding time from the original image database. Because this image data contains high-contrast artificial speckle texture to facilitate optical flow calculations, which can interfere with the identification of surface cracks, the data processing module 400 performs speckle removal processing on the retrieved image. The data processing module 400 uses morphological opening operations to filter out high-frequency speckle noise in the image, while preserving the low-frequency illumination characteristics and crack edge features of the gear surface itself. Let the original image be... , structural element is Image after speckle removal Represented as: ; in, Indicates an etching operation; This indicates an expansion operation. Based on this, the data processing module 400 pairs... The Canny edge detection operator is applied to extract the non-textured edges of the tooth surface.

[0063] The data processing module 400 identifies surface damage features, including at least tooth surface wear and tooth root cracks. If an irregular closed edge is detected in an abnormal area, it is determined to be tooth surface wear (such as peeling or pitting); if a linear extending edge is detected in the tooth root chamfer area, it is determined to be a tooth root crack. The identified defect type, location coordinates, and defect size (length, area) constitute the surface damage identification result.

[0064] The data processing module 400 integrates geometric parameter deviations and surface damage identification results to determine the health status and damage level of the gear. The health status and damage level of the gear are comprehensive evaluation conclusions derived from a preset fault classification strategy (e.g., "Level 1 / Normal", "Level 2 / Requires Observation", "Level 3 / Requires Return to Factory for Repair", "Level 4 / Scrapped"). For example, if there is only a slight geometric parameter deviation but no cracks, it is judged as "Level 2"; if there are root cracks, it is judged as "Level 4" regardless of whether the geometric parameters are out of tolerance.

[0065] After the inspection process is completed, the data processing module 400 automatically generates a standardized inspection report. The report includes the vehicle number (Vehicle number and wheelset number), inspection time, out-of-roundness value (i.e., the maximum radial runout value calculated above), a list of abnormal geometric parameters that exceed tolerances, and high-resolution images of the marked damaged areas.

[0066] The test report is uploaded to the vehicle operation and maintenance management system in real time via an industrial fieldbus (such as CAN bus or industrial Ethernet). The vehicle operation and maintenance management system refers to a host computer server software platform deployed in railway or rail transit depots for the management of all fleet assets, maintenance scheduling, and full lifecycle data monitoring. Simultaneously, a report summary is displayed on the on-site terminal for operators to review.

[0067] Finally, the data processing module 400 compresses and archives the raw acquired data and the processed structured data (including reconstructed contour coordinates, calculated geometric parameters, and judgment results) to the local storage unit. The local storage unit refers to a large-capacity solid-state drive or disk array inside the industrial control computer, used for long-term storage of historical data for subsequent historical tracing and trend analysis (such as analyzing the wear rate of the same gear in different maintenance cycles). The morphological operations, edge detection, and database query operations involved in the above data processing can be implemented using general programming techniques and algorithm libraries by those skilled in the art, and will not be elaborated upon here.

[0068] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An online detection system for gears in a rack train, characterized in that, include: A laser sensor array module includes at least two laser displacement sensors linearly distributed along the circumference of the gear rotation. The laser displacement sensors are used to continuously acquire vertical distance data of the gear tooth surface relative to the ground. The visual perception component module includes a structured light projector and an industrial camera. The structured light projector is used to project an artificial texture pattern onto the side of a gear, and the industrial camera is used to synchronously acquire a sequence of image data of the side of the gear containing the artificial texture pattern. The trigger module is used to monitor the position status of the vehicle wheelsets and send start and stop signals; The data processing module is communicatively connected to the laser sensor array module, the visual perception component module, and the trigger module. It is used to respond to the start signal to control the laser sensor array module and the visual perception component module to synchronously acquire and process the vertical distance data and the gear side image data sequence to obtain the gear's geometric contour and surface damage features.

2. The online detection system for gears in a rack train according to claim 1, characterized in that, In the laser sensor array module, the detection ranges of two adjacent laser displacement sensors in the circumferential direction of gear rotation physically overlap to form an overlapping field of view.

3. The online detection system for gears in a rack train according to claim 2, characterized in that, The data processing module is also used for: Select distance data segments from adjacent laser displacement sensors within the overlapping field of view; Perform waveform cross-correlation analysis on the distance data segment to separate the DC deviation caused by sensor installation error; The vertical distance data is corrected using the DC deviation.

4. The online detection system for gears in a rack train according to claim 1, characterized in that, The data processing module is also used for: An optical flow algorithm is applied to track the displacement features of the artificial texture pattern in the gear side image data sequence in order to calculate the global instantaneous angular velocity of the gear. Numerical integration is performed on the global instantaneous angular velocity to generate a sequence of gear rotation angles.

5. The online detection system for gears in a rack train according to claim 4, characterized in that, The data processing module is also used for: Establish a mapping relationship between the vertical distance data and the gear rotation angle sequence, and convert the vertical distance data in the time domain into angle domain data.

6. The online detection system for gears in a rack train according to claim 3, characterized in that, The data processing module is also used for: A digital low-pass filter is applied to the continuous distance data after the DC deviation has been eliminated to extract the AC vibration component caused by the vehicle suspension system and construct a vehicle vertical vibration data model. The vertical distance data is further corrected using the vehicle vertical vibration data output from the vehicle vertical vibration data model to obtain the corrected vertical distance data.

7. The online detection system for gears in a rack train according to claim 6, characterized in that, The data processing module is also used for: Based on the gear rotation angle sequence, the corrected vertical distance data is mapped to a two-dimensional polar coordinate system to reconstruct the full circumference geometric profile of the gear.

8. The online detection system for gears in a rack train according to claim 1, characterized in that, The data processing module is also used for: Based on the full circumferential geometric profile of the gear, the maximum radial runout value is calculated, and key geometric parameters, including tooth thickness and addendum circle diameter, are extracted. The maximum radial runout value is compared with the key geometric parameters and a preset reference parameter to obtain the geometric parameter deviation.

9. The online detection system for gears in a rack train according to claim 1, characterized in that, The data processing module is also used for: Identify regions in the full circumferential geometry of the gear that exhibit abrupt changes; Based on the angular index of the abnormal mutation region, the image data at the corresponding time is retrieved and speckle removal and edge detection are performed to identify surface damage features, including tooth surface wear and tooth root cracks.

10. An online inspection method for gears in a rack train, applied to the online inspection system for gears in a rack train as described in any one of claims 1-9, characterized in that, Includes the following steps: In response to the start signal sent by the trigger module, the vertical distance data of the gear tooth surface relative to the ground and the image data sequence of the gear side surface containing artificial texture patterns are collected synchronously. An optical flow algorithm is applied to track the gear side image data sequence, and the gear rotation angle sequence is calculated and integrated to generate the gear rotation angle sequence. Distance data segments from adjacent laser displacement sensors within the overlapping field of view are selected. DC deviation is separated through waveform cross-correlation analysis. The data with eliminated DC deviation is then low-pass filtered to extract AC vibration components. The DC deviation and AC vibration components are used to correct the vertical distance data to obtain corrected vertical distance data. Based on the gear rotation angle sequence, the corrected vertical distance data is mapped to a two-dimensional polar coordinate system to reconstruct the full circumference geometric profile of the gear; Based on the full circumference geometric contour of the gear, the geometric parameter deviation is calculated, and the surface damage characteristics are identified by analyzing the images corresponding to the abnormal change areas. Finally, the health status and damage level of the gear are comprehensively determined.

Images