A method for detecting and repairing defects of a railway rail by grinding

Abstract

The application discloses a kind of disease detection and polishing repair method of railway rail, including step 1: power on and start PLC, carry out polishing grinding wheel reset zero operation;Set the basic parameters of the rail to be polished this time, select the rail polishing area;Step 2: enter detection mode, execute disease identification, profile acquisition reconstruction, polishing parameter generation process, judge whether the rail needs polishing based on the rail profile detection result, if yes, execute step 3, the closed loop process of the application through defect identification, profile acquisition repair, parameter generation, accurate polishing, quality evaluation, through the step-by-step detection mode of fast scanning positioning defect and slow scanning acquisition profile, ensure the accuracy of polishing target;Combined with multi-curve fitting and profile alignment technology, ensure the accuracy of polishing amount calculation;Finally through standardized polishing operation and quality evaluation, realize the accurate repair of rail profile, realize the integrated operation of detection and polishing, improve the efficiency and safety of rail repair operation.

Description

Technical Field

[0001] This invention belongs to the field of railway rail repair and relates to a method for detecting and grinding repair defects in railway rails. Background Technology

[0002] During their service life, railway rails are constantly subjected to complex and variable loads from train wheels, making them prone to various damages and defects such as corrugation, cracks, spalling, crushing, and edge thickening. Without timely treatment and effective control, the service life of the rails will be significantly shortened, forcing premature rail replacement, wasting substantial resources, and increasing maintenance costs. More importantly, as the wheel-rail contact condition deteriorates, it severely affects train operation stability and passenger comfort, and may even lead to major safety accidents and casualties. Traditional methods for detecting rail defects suffer from low accuracy, inaccurate grinding, and difficulties in implementing grinding solutions. This application effectively solves the aforementioned problems. Summary of the Invention

[0003] In order to overcome at least one deficiency of the prior art, the present invention provides a method for detecting and grinding repair defects in railway rails.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: a method for detecting and grinding repair defects in railway rails, comprising the following steps: Step 1: Connect the power supply and start the PLC to perform the grinding wheel reset and zeroing operation; set the basic parameters of the rail to be ground and select the rail grinding area; Step 2: Enter detection mode and execute the process of defect identification, profile acquisition and reconstruction, and grinding parameter generation. Based on the rail profile detection results, determine whether the rail needs grinding. If yes, proceed to step 3; otherwise, end the process. Step 3: Enter grinding mode, and use the grinding wheel to perform the rail grinding process until the rail profile meets the standard after grinding; Step 4: Reset the grinding wheel to zero to complete the grinding and repair of this section of rail; Step 5: Construct a comprehensive evaluation system for rail repair quality, and evaluate the rail repair quality online based on the comprehensive evaluation system; Step 6: End of steps.

[0005] Furthermore, the method for disease identification in step 2 includes... Step A1: Establish a visual feature library of normal images of rail surfaces and a visual feature library of defective images of rail surfaces; Step A2: The grinding trolley moves quickly along the inspected rail section and captures images of the rail surface using a vision camera; Step A3: Obtain the visual features of the rail surface image; Step A4: Compare the visual features of the image with the visual feature library of normal images to complete the initial screening of normal track surfaces and suspected defect areas, remove images of areas without defects, and retain images of suspected defect areas; Step A5: Call the defect feature library of the track surface and identify the suspected area image based on template matching; Step A6: Using a machine learning model, classify the identified defects according to the classification criteria and record the location of the defect areas.

[0006] Furthermore, the contour acquisition and reconstruction method in step 2 is as follows: Step B1: Perform initial calibration on the standard gauge blocks and establish the initial transformation relationship between the two profilers and the global coordinate system; Step B2: The grinding trolley moves slowly in the reverse direction to the defect area. The profiler scans the rail profile of the defect area to obtain multiple rail profile samples and profile data. Step B3: Calculate the statistical deviation of the double profiler data in the overlapping area in real time, and correct the conversion coefficient based on the statistical deviation to ensure that the coordinates of the same physical point are consistent after the double profiler conversion, without relying on any standard rail features. Step B4: Verify the transformation coefficients and calculate the coefficient of variation; Step B5: Reconstruct the complete rail surface data based on the profile data stitching; Step B6: Based on the complete rail surface data, obtain the representative profile of the rail in the defect area.

[0007] Furthermore, the method for generating the polishing parameters in step 2 is as follows: Step C1: Use the standard rail profile of this section of the line as the target profile for grinding; Step C2: Compare and analyze the representative profile of the rail with the target profile for grinding, and calculate the amount of grinding required; Step C3: Based on the characteristics of the grinding device, formulate grinding parameters and design a grinding strategy.

[0008] Furthermore, the rail grinding process in step 3 includes... Step 30: Construct a grinding quality control mechanism based on multiple corrections, and control the grinding quality of rails based on the grinding quality control mechanism; Step 31: Based on the grinding strategy, select the grinding method. If manual grinding is required, execute the manual grinding process. If automatic grinding is required, execute the automatic grinding process to complete the rail grinding operation. Step 32: The profiler scans the profile of the polished rail to obtain multiple rail profile samples and profile data, and reconstructs the complete rail surface data after polishing by stitching together the profile data; based on the complete rail surface data, a representative profile of the polished rail is obtained; Step 33: Compare and analyze the representative profile of the polished rail with the target profile, calculate the deviation of the polished rail profile, compare the deviation of the rail profile with the set deviation threshold, and determine whether the deviation of the rail profile is less than the deviation threshold. If so, it is considered that the polished rail profile has reached the target profile; otherwise, polish again until the target profile is reached.

[0009] Furthermore, the method for constructing a grinding quality control mechanism based on multiple corrections in step 30 is as follows: Step 301: Construct the first layer of correction: passive anti-deviation walking profile limit; The grinding carriage is equipped with traveling wheels, which roll on the steel rails. When the grinding carriage moves, the traveling wheels roll on the steel rails; the outer contour of the traveling wheels matches the profile of the steel rail surface. Step 302: Construct the second layer of correction: passively reinforced anti-deviation travel limit mechanism with rail clamping limit; The grinding trolley is equipped with a travel limit mechanism. The limit wheels of the travel limit mechanism are arranged in a symmetrical rail-hugging layout, and the inner side of the limit wheels is in contact with the lower part of the rail head. Step 303: Construct the third layer of correction: dynamic feed compensation of the ranging sensor to actively compensate for deviation; Step 304: Construct the fourth level of correction: active pressure stabilization pressure sensor constant force control; Step 305: Set up protection mechanism When F real >1.2F target It was determined that a sudden pressure surge exceeded the limit, triggering an emergency feed retraction while maintaining vehicle movement to prevent overload damage to the grinding wheel; F real <0.8F target When the set time is continuously reached, grinding is paused, a non-contact warning is issued, and the condition of the rail surface or grinding wheel is checked. Actual feed depth calculation: in, For longitudinal feed rate based on target profile; This is the final longitudinal feed amount.

[0010] Furthermore, the method of step 303 is as follows: The laser rangefinder sensor detects the non-working edge of the rail head and measures the horizontal distance between the sensor and the non-working edge in real time, denoted as L. real Preset standard distance L std When the vehicle moves laterally to the left or right, L real With L std The deviation distance ΔL is generated: ΔL=L real -Lstd ; This deviation corresponds to the positional deviation between the grinding wheel and the working edge of the rail head. This deviation is converted into a feed depth compensation amount Δxs using an algorithm, and the grinding wheel feed is adjusted in real time to counteract the lateral movement. Lateral deviation conversion: ΔL and feed depth compensation Δxs satisfy a linear relationship: Where k is the compensation coefficient, and there is an installation angle θ between the grinding wheel and the sensor, k=cosθ; Actual transverse feed calculation: Where, x base x is the lateral feed rate based on the target profile. actual This is the final lateral feed amount. When ΔL>ΔLmax, ΔLmax is the mechanical limit, the system issues a warning, reduces the vehicle's travel speed, and avoids excessive compensation range leading to grinding deviation; at the same time, when collecting distance data through the distance sensor, the acquired data is processed by median filtering to eliminate measurement noise caused by rail surface contamination.

[0011] Furthermore, the automated polishing process includes Step E1: Adjust the grinding wheel to the current grinding angle according to the grinding angle adjustment requirements; the system automatically controls the grinding wheel to perform lifting and lateral movements for tool setting. After tool setting is completed, the number of grinding passes is automatically recorded, the set feed parameters are read, and the transverse and longitudinal feeds are completed. Step E2: Read the set walking speed parameters and control the walking mechanism to complete the movement of the grinding carriage; Step E3: Based on the distance measurement and pressure values ​​fed back during the walking process, adjust the lateral and longitudinal power mechanisms in real time to achieve feedback control; Step E4: When the distance of the grinding carriage reaches the set value, stop moving and determine whether the current number of grinding passes has reached the set value; if yes, proceed to step E5; if no, accumulate the number of grinding passes, move the grinding carriage in the opposite direction, and proceed to step E3. Step E5: Determine whether to traverse the grinding angles. If yes, end the step; otherwise, proceed to step E1.

[0012] Furthermore, step 304 achieves pressure stabilization by adjusting the feed depth, as shown in the following formula. Where Δz is the feed longitudinal depth adjustment based on pressure deviation; K p For proportionality coefficient, K i For the integral coefficient, K dare the differential coefficients, the three coefficients are constants, and n is the current sampling time.

[0013] Furthermore, the formula for the Comprehensive Quality Evaluation Index (CGQI) of rail grinding in step 5 is as follows: in, , , , For dynamic weights, DPA is the dynamic profile fit; DRC is the defect residue correction degree; SRC is the surface roughness compliance degree; GUI is the grinding quality uniformity score.

[0014] In summary, the advantages of this invention are: This invention employs a closed-loop process encompassing defect identification, profile acquisition and repair, parameter generation, precise grinding, and quality evaluation. It utilizes a step-by-step detection mode that combines rapid scanning to locate defects with slow scanning to acquire profiles, ensuring precise grinding targets. Furthermore, it incorporates multi-curve fitting and profile alignment technologies to guarantee the accuracy of grinding quantity calculations. Finally, through standardized grinding operations and quality evaluation, it achieves precise repair of the rail profile, realizing integrated inspection and grinding operations, and improving the efficiency and safety of rail repair work. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0016] Figure 2 This is a flowchart of the profile acquisition and reconstruction process of the present invention.

[0017] Figure 3 This is a schematic diagram of the profiler installation according to the present invention.

[0018] Figure 4 This is a schematic diagram of the track surface imaging of the present invention.

[0019] Figure 5 This is a schematic diagram of the automatic polishing process of the present invention.

[0020] Figure 6 This is a flowchart of the multiple correction mechanism of the present invention. Detailed Implementation

[0021] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0022] Example: like Figures 1-6 As shown, a method for detecting and repairing defects in railway rails includes the following steps: Step 1: Connect the power supply and start the PLC to perform the grinding wheel reset and zeroing operation; set the basic parameters of the rail to be ground and select the rail grinding area; Step 2: Enter detection mode and execute the process of defect identification, profile acquisition and reconstruction, and grinding parameter generation. Based on the rail profile detection results, determine whether the rail needs grinding. If yes, proceed to step 3; otherwise, end the process. Step 3: Enter grinding mode, and use the grinding wheel to perform the rail grinding process until the rail profile meets the standard after grinding; Step 4: Reset the grinding wheel to zero to complete the grinding and repair of this section of rail; Step 5: Construct a comprehensive evaluation system for rail repair quality, and evaluate the rail repair quality online based on the comprehensive evaluation system; Step 6: End of steps.

[0023] The methods for disease identification in step 2 include: Step A1: Establish a visual feature library of normal images of rail surfaces and a visual feature library of defective images of rail surfaces; Image visual features include grayscale differences, texture abrupt changes, and geometric shapes.

[0024] Step A2: The grinding trolley moves quickly along the inspected rail section and captures images of the rail surface using a vision camera; Step A3: Obtain the visual features of the rail surface image; Step A4: Compare the visual features of the image with the visual feature library of normal images to complete the initial screening of normal track surfaces and suspected defect areas, remove images of areas without defects, and retain images of suspected defect areas; Step A5: Call the defect feature library of the track surface and identify the suspected area image based on template matching; Step A6: Using a machine learning model, classify the identified defects according to the classification criteria and record the location of the defect areas.

[0025] The contour acquisition and reconstruction method in step 2 is as follows: Step B1: Perform initial calibration on the standard gauge blocks and establish the initial transformation relationship between the two profilers and the global coordinate system; There are two sets of profile measuring instruments, which are installed on both sides of the track to obtain rail profile data, thus enabling complete reading of all rail surface information.

[0026] Before automating the equipment operation, initial calibration should be performed. Place a standard gauge block containing at least three marker points in the overlapping area scanned by both profilers (e.g., X∈[-5,5]mm). Taking the calculation of three marker points as an example, ensure that both profilers can completely scan the three marker points of the gauge block.

[0027] For ease of description, the two sets of profile gauges are named: Profile Gauge #1 and Profile Gauge #2, and the three marker points are labeled as 1, 2, and 3. The profiler #1 collected the local coordinates of three marker points: Q1 represents the local coordinate set of three marker points collected by profiler #1. , , The local coordinates of the three marker points collected by profiler #1; The profiler #2 collected the local coordinates of three marker points: Q2 represents the local coordinate set of three marker points collected by profiler #2. , , The local coordinates of the three marker points collected by profiler #2; Given the global coordinates of three marker points: Among them, Q global The global coordinate set of the three marker points. , , The global coordinates of the three marker points; For each profiler, the coordinate transformation formula is fitted using the least squares method; in, , , Let be the rotation angle, and c and d be the translation amounts.

[0028] The objective function obtained by the least squares method is: The coordinate transformation formula obtained from solving the problem using profiler #1 is: The coordinate transformation formula obtained from the solution obtained by profiler #2 is: Step B2: The grinding trolley moves slowly in the reverse direction to the defect area. The profiler scans the rail profile of the defect area to obtain multiple rail profile samples and profile data. Step B3: Calculate the statistical deviation of the double profiler data in the overlapping area in real time, and correct the conversion coefficient based on the statistical deviation to ensure that the coordinates of the same physical point are consistent after the double profiler conversion, without relying on any standard rail features. The overlapping area (rail top core area) of profile gauges #1 and #2 measures the same physical area. Regardless of whether the rail is worn, the data deviation of the two instruments in this area directly reflects the drift of the coordinate transformation relationship.

[0029] This step uses data from the overlapping area to correct the transformation relationship in real time, thus offsetting operational drift. Within the overlapping area, the relevant parameters of the #1 profiler are corrected based on the coordinates of the #2 profiler. Therefore, only the relevant parameters corresponding to the coordinate transformation formula of the #1 profiler need to be corrected; the coordinate transformation formula of the #2 profiler remains unchanged.

[0030] Step B31: Collect overlapping area data; The profiler #1 collects the point set P1 of the overlapping area of ​​the rails, and the profiler #2 collects the point set P2 of the overlapping area of ​​the rails. Step B32: Extract overlapping area data; Use transformation formulas T1 and T2 to convert P1 and P2 to global coordinates respectively. , ; Extracting overlapping region data: Step B33: Alignment and deviation calculation of overlapping area data; by Based on the X coordinate, for Perform linear interpolation to obtain the Y value at the same X position.

[0031] Calculate the Y deviation at each X position: Where k = 1, 2, ..., t; t is the number of coordinates of the overlapping region; for Y coordinate, for The Y-coordinate.

[0032] by Based on the Y coordinate, for Perform linear interpolation to obtain the X value at the same Y position.

[0033] Calculate the positional deviation for each X position: Where k = 1, 2, ..., t; t is the number of coordinates of the overlapping region; for The X coordinate, for The X coordinate.

[0034] Step B34: Calculate the mean deviation; Step B35: Correct the conversion factor; Assuming the rotation coefficient is fixed, correct the translation coefficient of the coordinate transformation formula for profiler #1: in, The Y-position correction translation amount for the coordinate transformation formula of profiler #1; This represents the Y-position translation amount in the coordinate transformation formula for profiler #1; The X-position correction translation amount for the coordinate transformation formula of profiler #1; This represents the X-position translation amount in the coordinate transformation formula of profiler #1.

[0035] Step B36: Smooth Update; in, For smoothing coefficients, ; Adjust the translation amount for the last X position. Adjust the translation amount for the last Y position; To correct the translation amount for the updated X position, Adjust the translation amount for the updated Y position.

[0036] Step B4: Verify the transformation coefficients and calculate the coefficient of variation; Three adjacent cross-sections were continuously sampled (0.5m apart), and the corrected translation amount for each cross-section was calculated. as well as ; Calculate the coefficient of variation (CV): CV = (Standard deviation / Mean) * 100% Determine if CV is ≤3%. If so, take the average of the three section coefficients as the final coordinate transformation coefficient of the current measurement track surface. as well as If not, proceed to step B2 and continue iterating.

[0037] The final coordinate transformation formula for profiler #1 is: The coordinate transformation formula for profiler #2 is: Step B5: Reconstruct the complete rail surface data based on the profile data stitching; Full coordinate transformation: Using the modified transformation formula, all the coordinates collected by profilers #1 and #2 are converted to global coordinates. as well as Overlapping region fusion: For points located in the overlapping region X∈[-5,5] mm, take (Weighted average, eliminating measurement noise); where, The Y-coordinate calculated for the overlapping area. The Y-coordinate of profiler #1 in the overlapping area. The Y-coordinate of profiler #2 in the overlapping area.

[0038] Merging non-overlapping regions: (X>5mm) and (X < 5mm) Merge directly; Sorting output: Sort by X coordinate from smallest to largest to obtain a complete set of discrete points for the rail profile, thus completing the reconstruction of the complete rail surface data.

[0039] Step B6: Based on complete rail surface data, obtain the representative profile of the rail in the defect area; Based on complete rail surface data, a profile curve is fitted using a multi-curve fitting algorithm. The shape wear characteristics are then compared and analyzed with all measured profile samples. If the shape wear comparison analysis is satisfactory, the fitted profile curve is determined to be the most representative profile curve that matches the sample profile, and it is used as the representative profile of the rail in that area. If the shape wear comparison analysis is not satisfactory, the profile curve is refitted until the shape wear characteristics comparison analysis is satisfied.

[0040] The method for generating grinding parameters in step 2 is as follows: Step C1: Use the standard rail profile of this section of the line as the target profile for grinding; Step C2: Compare and analyze the representative profile of the rail with the target profile for grinding, and calculate the amount of grinding required; Step C3: Based on the characteristics of the grinding device, formulate grinding parameters and design a grinding strategy; Grinding parameters include grinding method, number of grinding passes, grinding power, grinding speed, grinding angle parameters; and the grinding angle adjustment angle of the grinding wheel.

[0041] Step 3, the rail grinding process, includes: Step 30: Construct a grinding quality control mechanism based on multiple corrections, and control the grinding quality of rails based on the grinding quality control mechanism; Step 31: Based on the grinding strategy, select the grinding method. If manual grinding is required, execute the manual grinding process. If automatic grinding is required, execute the automatic grinding process to complete the rail grinding operation. Step 32: The profiler scans the profile of the polished rail to obtain multiple rail profile samples and profile data, and reconstructs the complete rail surface data after polishing by stitching together the profile data; based on the complete rail surface data, a representative profile of the polished rail is obtained; The specific operation is the same as step 2, and will not be repeated here; Step 33: Compare and analyze the representative profile of the polished rail with the target profile, calculate the deviation of the polished rail profile, compare the deviation of the rail profile with the set deviation threshold, and determine whether the deviation of the rail profile is less than the deviation threshold. If so, it is considered that the polished rail profile has reached the target profile; otherwise, polish again until the target profile is reached.

[0042] The manual polishing process includes Step D1: Manually adjust the grinding wheel to the current grinding angle according to the grinding angle; manually control the grinding wheel to perform lifting and lateral movements to complete tool setting, and set the grinding amount; Step D2: Manually feed the grinding wheel and grind the rail at a specific deflection angle. Then continue to adjust the deflection angle and repeat the above operation process to complete the grinding of the rail in this area.

[0043] The automatic polishing process includes Step E1: Adjust the grinding wheel to the current grinding angle according to the grinding angle adjustment requirements; the system automatically controls the grinding wheel to perform lifting and lateral movements for tool setting. After tool setting is completed, the number of grinding passes is automatically recorded, the set feed parameters are read, and the transverse and longitudinal feeds are completed. Step E2: Read the set walking speed parameters and control the walking mechanism to complete the movement of the grinding carriage; Step E3: Based on the distance measurement and pressure values ​​fed back during the walking process, adjust the lateral and longitudinal power mechanisms in real time to achieve feedback control; Step E4: When the distance of the grinding carriage reaches the set value, stop moving and determine whether the current number of grinding passes has reached the set value; if yes, proceed to step E5; if no, accumulate the number of grinding passes, move the grinding carriage in the opposite direction, and proceed to step E3. Step E5: Determine whether to traverse the grinding angles. If yes, end the step; otherwise, proceed to step E1.

[0044] Step 30 describes the method for constructing a grinding quality control mechanism based on multiple corrections: Step 301: Constructing the first layer of correction: Passive anti-deviation walking profile limiter: The grinding carriage is equipped with traveling wheels, which roll on steel rails. When the grinding carriage moves, the traveling wheels roll on the steel rails.

[0045] The outer contour of the traveling wheels conforms to the profile of the rail surface (taking a 60kg / m rail as an example, with a rail top curvature R=300mm, the curvature of the traveling wheel contact surface is consistent with the rail top). The curved surface of the traveling wheels and the rail surface forms a natural limit, restricting the lateral movement of the vehicle body along the width direction of the rail. The traveling wheels are made of wear-resistant alloy steel (hardness HRC55~60), and the contact surface has a fish-scale pattern to enhance the friction with the rail surface.

[0046] Step 302: Construct the second layer of correction: passively reinforced anti-deviation travel limit mechanism with rail clamping limit; The grinding trolley is equipped with a fixed travel limiting mechanism. The limiting wheels of the mechanism have a symmetrical rail-hugging layout, with the inner side of the limiting wheels tightly fitted to the lower part of the rail head. The limiting wheels adopt an elastic suspension structure (spring stiffness k=50N / mm), which can adapt to the slight deformation of the rail and avoid vehicle vibration caused by rigid contact. The limiting wheels further restrict the lateral movement of the vehicle body through the elastic rail-hugging mechanism, further compressing the residual lateral movement after the first correction to the millimeter level. The limiting wheels are made of polyurethane (Shore hardness A90).

[0047] Step 303: Construct the third layer of correction: dynamic feed compensation of the ranging sensor to actively compensate for deviation; The laser rangefinder has a measurement range of -15 to +15 mm and an accuracy of ±0.01 mm. It is installed on one side of the vehicle body.

[0048] The laser rangefinder sensor detects the non-working edge of the rail head (because its deformation is negligible, it can be used as a reference). It measures the horizontal distance between the rangefinder sensor and the non-working edge of the rail head in real time, denoted as L. real Preset standard distance L std (The distance is calculated based on the target profile after the sensor installation position is fixed, i.e., the theoretical distance without lateral movement); when the vehicle moves laterally, L real With L std The deviation distance ΔL is generated: ΔL=L real -L std ; This deviation directly corresponds to the positional deviation between the grinding wheel and the working edge of the rail head. It is converted into a feed depth compensation amount Δxs through an algorithm, and the feed of the grinding wheel is adjusted in real time to offset the lateral movement.

[0049] Lateral deviation conversion: ΔL and feed depth compensation Δx satisfy a linear relationship (because the installation positions of the grinding wheel and sensor are fixed, the geometric relationship is determined): Where k is the compensation coefficient, which is calibrated by the mechanical structure of the equipment. Since there is an installation angle θ between the grinding wheel and the sensor, k = cosθ. Actual transverse feed calculation: Where, x base x is the lateral feed rate based on the target profile (derived from theoretical calculations); actual This is the final lateral feed amount.

[0050] When ΔL > ΔLmax, ΔLmax is mechanically limited, the system issues an early warning, and at the same time reduces the vehicle's travel speed to avoid excessive compensation range leading to grinding deviation; at the same time, when collecting distance data through the distance sensor, the acquired data is processed by median filtering to eliminate measurement noise caused by rail surface contamination (such as rust).

[0051] Step 304: Construct the fourth level of correction: active pressure stabilization pressure sensor constant force control; The pressure sensor is a column-type pressure sensor (measuring range 0~500N, accuracy ±1%), installed in the longitudinal module of the grinding wheel. The longitudinal module controls the longitudinal movement of the grinding wheel and collects the contact pressure F between the grinding wheel and the rail in real time. real Based on the grinding strategy, a preset target pressure F is established. target Perform constant force control during grinding.

[0052] When hard spots appear in the grinding area, the rail surface is uneven, or the grinding wheel is worn, F real It will deviate from F target By adjusting the feed depth through closed-loop control, F real Stable at F target Within ±5%, avoid excessive pressure leading to over-polishing and insufficient pressure leading to under-polishing.

[0053] This application employs an incremental PID algorithm to achieve pressure stability by adjusting the feed depth, as shown in the following formula: Where Δz is the feed longitudinal depth adjustment based on pressure deviation; K p For proportionality coefficient, K i For the integral coefficient, K d These are the differential coefficients; the three coefficients are constants and can be adjusted according to actual working conditions; n is the current sampling time.

[0054] Step 305: Set up a protection mechanism; When F real >1.2F target It was determined that there was a sudden pressure surge exceeding the limit, triggering an emergency feed retraction in the system while maintaining vehicle movement to prevent overload damage to the grinding wheel; F real <0.8F target If the grinding continues for a certain period of time (which can be set to 2 seconds), the system will pause grinding, issue a non-contact warning, and check the condition of the rail surface or grinding wheel.

[0055] Actual feed depth calculation: in, The longitudinal feed rate is based on the target profile (calculated theoretically). This is the final longitudinal feed amount.

[0056] The method for constructing the comprehensive quality evaluation system (CGQI) for rail repair in step 5 includes: Step 51: Calculate the dynamic profile fit (DPA); The steps for calculating the Dynamic Profile Fit (DPA) include: Step 511: Obtain basic wheel and rail parameters; The basic parameters of the wheel and rail system are shown in Table 1: Table 1 Step 512: Divide the rail head width direction into n infinitesimal regions Z according to the infinitesimal width Δf. i ; The width of the infinitesimal element is Δf = 0.5 mm, and the region of the infinitesimal element is Z. i (i=1,2,...,n), infinitesimal region Z i Covers the core area of ​​wheel-rail contact (center of rail top ±25mm); Step 513: Calculate the contact pressure P in each micro-element region i ; Contact pressure P in each micro-element region i The formula is: Where, x i For the infinitesimal element Z i Center x-axis coordinate, yi For the infinitesimal element Z i The central y-axis coordinate, y i Very small, remember y i =0; P i The formula has been updated to in, For maximum contact pressure, h is the major semi-axis of the contact ellipse (along the width direction of the rail head). j is the minor semi-axis of the contact ellipse (along the length of the rail). R is the equivalent radius of curvature. ; For the equivalent elastic modulus, .

[0057] Step 514: Weight the contact pressure Normalization processing; The greater the contact pressure, the higher the weight, reflecting the dynamic evaluation logic. Step 515: Calculate the normal deviation based on the profiler measurement results. ; Profile alignment: Using the track gauge point (16mm below the highest point of the track top) as a reference, translate the profile measured by the laser profile measuring instrument along the Y-axis to align it with the target profile; Deviation calculation: for each infinitesimal element Calculate the shortest normal distance between the measured profile and the target profile, and take the absolute value as... ; Deviation handling: If > ,according to = Handling (to avoid excessive deviations that could lead to distorted evaluations). This is the limit for normal deviation.

[0058] Step 516: Calculate the dynamic profile fit degree (DPA); DPA ∈ [0,100], the higher the score, the better the dynamic fit of the profile.

[0059] Step 52: Calculate the disease residue correction factor (DRC); Set the severity level of the disease: Fatal defects (L1): Cracks (length ≥ 3 mm), spalling (area ≥ 5 mm²), correction factor =0.3; Severe disease (L2): Excessive lip growth (height ≥ 0.5 mm), peeling (area ≥ 10 mm²), correction factor. =0.7; Minor damage (L3): minor scratches, fish-scale pattern (depth <0.2mm), correction factor. =0.95; DRC formula for disease residue correction: Where m is the number of residual disease types; if there are 2 disease types, then m=2. is the correction factor for the j-th type of disease. When there are no residual diseases, DRC=100.

[0060] Step 53: Calculate the surface roughness compliance (SRC); The surface roughness compliance (SRC) formula is: in, The surface roughness is measured (one measuring point is taken every 1 meter, and the average value is taken; high-speed rail). Ordinary speed rail ); For optimal roughness (midpoint of the threshold interval, such as high-speed rail) = 2.0 ), To minimize roughness, This represents the maximum roughness.

[0061] Step 54: Calculate the uniformity of polishing quality (GUI); The quality of grinding depends not only on whether the profile of a single section meets the standard (DPA has been evaluated), but also on whether the grinding amount is uniform throughout the entire rail section. Even if a single section has a high DPA score, large differences in grinding amount between adjacent sections can still lead to unstable wheel-rail contact and shortened rail service life. The coefficient of variation (CV) is used to eliminate the influence of the average grinding amount and quantify the dispersion of the grinding amount. A maximum allowable coefficient of variation (CV) is set. lim The score is calculated by the ratio of the actual coefficient of variation to the allowable value, thus realizing the evaluation logic that the better the uniformity, the higher the score.

[0062] Step 541: Calculate the grinding amount t for a single cross-section k ; The grinding amount is the deviation area between the measured profile and the target profile, representing the total amount of material that needs to be ground away from that section. The formula is: Among them, t k Grinding amount for the k-th cross-section (unit: mm) 2), k=1,2,...,n1 (n1 is the number of cross-sections, usually one cross-section is taken per meter), The target profile function (x is the coordinate of the railhead width direction, and f(x) is the corresponding height); The measured profile function of the cross-section (data acquired by a laser profiler and aligned with the track gauge points); integration To account for the area being ground (rather than single-point deviation), the deviation values ​​of all infinitesimal elements along the rail head width direction need to be summed by integration. The laser profiler will output the coordinates of discrete points along the rail head width direction (e.g., one point every 0.1 mm), and the integral can be approximated using the trapezoidal rule. in, Let be the width coordinate of the i-th discrete point, where i = 1, 2, 3, ..., mm, and mm is the number of discrete points within the grinding area. The discrete point spacing (sampling interval of the laser profiler, here taken as 0.1mm); Step 542: Calculate the coefficient of variation (CV(t)) of the grinding amount. k ); CV(t k The ratio of standard deviation to mean is used to eliminate the influence of the mean value of grinding amount and objectively reflect the degree of dispersion. in, This represents the average amount of polishing. ; For the standard deviation of polishing amount, .

[0063] Step 543: Calculate the GUI; The GUI provides a score for the uniformity of the polishing quality (range 0-100 points, the higher the score, the better the uniformity). The maximum allowable coefficient of variation for achieving uniformity in polishing. If CV(t) k If CV(t) = 0 (the polishing amount is completely uniform), then GUI = 100 points; if CV(t) = 0 (the polishing amount is completely uniform), then GUI = 100 points; k )=CV lim (If the maximum allowable dispersion is reached), then GUI = 0 points; if it is between these two values, the score decreases linearly; when CV(t) k )>CV lim When the time is right, GUI score is set to 0.

[0064] Step 55: Calculate the Comprehensive Quality Evaluation Index (CGQI) for rail grinding. in, , , , The weighting is dynamic, adjusted according to the line scenario. Priority is given to increasing the DPA weight for critical areas (such as turnouts) and high-speed lines. The specific allocation is shown in Table 2. Table 2 CGQI ∈ [0,100], the higher the score, the better the polishing quality. The correlation between CGQI score, quality level, and recommendations is shown in Table 3: This application employs a closed-loop process encompassing defect identification, profile acquisition and repair, parameter generation, precision grinding, and quality evaluation. It utilizes a step-by-step detection mode that combines rapid scanning to locate defects with slow scanning to acquire profiles, ensuring precise grinding targets. Furthermore, it combines multi-curve fitting and profile alignment technologies to guarantee the accuracy of grinding quantity calculations. Finally, through standardized grinding operations and quality evaluation, it achieves precise repair of the rail profile.

[0065] Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

Claims

1. A method for detecting and grinding repair defects in railway rails, characterized in that: Includes the following steps: Step 1: Connect the power supply and start the PLC to perform the grinding wheel reset and zeroing operation; set the basic parameters of the rail to be ground and select the rail grinding area; Step 2: Enter detection mode and execute the process of defect identification, profile acquisition and reconstruction, and grinding parameter generation. Based on the rail profile detection results, determine whether the rail needs grinding. If yes, proceed to step 3; otherwise, end the process. Step 3: Enter grinding mode, and use the grinding wheel to perform the rail grinding process until the rail profile meets the standard after grinding; Step 4: Reset the grinding wheel to zero to complete the grinding and repair of this section of rail; Step 5: Construct a comprehensive evaluation system for rail repair quality, and evaluate the rail repair quality online based on the comprehensive evaluation system; Step 6: End of steps.

2. The method for detecting and grinding repair defects in railway rails according to claim 1, characterized in that: The method for disease identification in step 2 includes: Step A1: Establish a visual feature library of normal images of rail surfaces and a visual feature library of defective images of rail surfaces; Step A2: The grinding trolley moves quickly along the inspected rail section and captures images of the rail surface using a vision camera; Step A3: Obtain the visual features of the rail surface image; Step A4: Compare the visual features of the image with the visual feature library of normal images to complete the initial screening of normal track surfaces and suspected defect areas, remove images of areas without defects, and retain images of suspected defect areas; Step A5: Call the defect feature library of the track surface and identify the suspected area image based on template matching; Step A6: Using a machine learning model, classify the identified defects according to the classification criteria and record the location of the defect areas.

3. The method for detecting and grinding repair defects in railway rails according to claim 1, characterized in that: The contour acquisition and reconstruction method in step 2 is as follows: Step B1: Perform initial calibration on the standard gauge blocks and establish the initial transformation relationship between the two profilers and the global coordinate system; Step B2: The grinding trolley moves slowly in the reverse direction to the defect area. The profiler scans the rail profile of the defect area to obtain multiple rail profile samples and profile data. Step B3: Calculate the statistical deviation of the double profiler data in the overlapping area in real time, and correct the conversion coefficient based on the statistical deviation to ensure that the coordinates of the same physical point are consistent after the double profiler conversion, without relying on any standard rail features. Step B4: Verify the transformation coefficients and calculate the coefficient of variation; Step B5: Reconstruct the complete rail surface data based on the profile data stitching; Step B6: Based on the complete rail surface data, obtain the representative profile of the rail in the defect area.

4. The method for detecting and grinding repair defects in railway rails according to claim 1, characterized in that: The method for generating grinding parameters in step 2 is as follows: Step C1: Use the standard rail profile of this section of the line as the target profile for grinding; Step C2: Compare and analyze the representative profile of the rail with the target profile for grinding, and calculate the amount of grinding required; Step C3: Based on the characteristics of the grinding device, formulate grinding parameters and design a grinding strategy.

5. The method for detecting and grinding repair defects in railway rails according to claim 1, characterized in that: The rail grinding process in step 3 includes: Step 30: Construct a grinding quality control mechanism based on multiple corrections, and control the grinding quality of rails based on the grinding quality control mechanism; Step 31: Based on the grinding strategy, select the grinding method. If manual grinding is required, execute the manual grinding process. If automatic grinding is required, execute the automatic grinding process to complete the rail grinding operation. Step 32: The profiler scans the profile of the polished rail to obtain multiple rail profile samples and profile data, and reconstructs the complete rail surface data after polishing by stitching together the profile data; based on the complete rail surface data, a representative profile of the polished rail is obtained; Step 33: Compare and analyze the representative profile of the polished rail with the target profile, calculate the deviation of the polished rail profile, compare the deviation of the rail profile with the set deviation threshold, and determine whether the deviation of the rail profile is less than the deviation threshold. If so, it is considered that the polished rail profile has reached the target profile; otherwise, polish again until the target profile is reached.

6. The method for detecting and grinding repair defects in railway rails according to claim 5, characterized in that: The method for constructing a grinding quality control mechanism based on multiple corrections in step 30 is as follows: Step 301: Construct the first layer of correction: passive anti-deviation walking profile limit; The grinding carriage is equipped with traveling wheels, which roll on the steel rails. When the grinding carriage moves, the traveling wheels roll on the steel rails; the outer contour of the traveling wheels matches the profile of the steel rail surface. Step 302: Construct the second layer of correction: passively reinforced anti-deviation travel limit mechanism with rail clamping limit; The grinding trolley is equipped with a travel limit mechanism. The limit wheels of the travel limit mechanism are arranged in a symmetrical rail-hugging layout, and the inner side of the limit wheels is in contact with the lower part of the rail head. Step 303: Construct the third layer of correction: dynamic feed compensation of the ranging sensor to actively compensate for deviation; Step 304: Construct the fourth level of correction: active pressure stabilization pressure sensor constant force control; Step 305: Set up protection mechanism When F real >1.2F target It was determined that a sudden pressure surge exceeded the limit, triggering an emergency feed retraction while maintaining vehicle movement to prevent overload damage to the grinding wheel; F real <0.8F target When the set time is continuously reached, grinding is paused, a non-contact warning is issued, and the condition of the rail surface or grinding wheel is checked. Actual feed depth calculation: ,in, For longitudinal feed rate based on target profile; This is the final longitudinal feed amount.

7. The method for detecting and grinding repair defects in railway rails according to claim 6, characterized in that: The method of step 303 is as follows: The laser rangefinder sensor detects the non-working edge of the rail head and measures the horizontal distance between the sensor and the non-working edge in real time, denoted as L. real Preset standard distance L std When the vehicle moves laterally to the left or right, L real With L std The deviation distance ΔL is generated: ΔL=L real -L std ; This deviation corresponds to the positional deviation between the grinding wheel and the working edge of the rail head. This deviation is converted into a feed depth compensation amount Δxs using an algorithm, and the grinding wheel feed is adjusted in real time to counteract the lateral movement. Lateral deviation conversion: ΔL and feed depth compensation Δxs satisfy a linear relationship: Where k is the compensation coefficient, and there is an installation angle θ between the grinding wheel and the sensor, k=cosθ; Actual transverse feed calculation: , where x base x is the lateral feed rate based on the target profile. actual This is the final lateral feed amount. When ΔL>ΔLmax, ΔLmax is the mechanical limit, the system issues a warning, reduces the vehicle's travel speed, and avoids excessive compensation range leading to grinding deviation; at the same time, when collecting distance data through the distance sensor, the acquired data is processed by median filtering to eliminate measurement noise caused by rail surface contamination.

8. The method for detecting and grinding repair defects in railway rails according to claim 5, characterized in that: The automatic polishing process includes Step E1: Adjust the grinding wheel to the current grinding angle according to the grinding angle adjustment requirements; the system automatically controls the grinding wheel to perform lifting and lateral movements for tool setting. After tool setting is completed, the number of grinding passes is automatically recorded, the set feed parameters are read, and the transverse and longitudinal feeds are completed. Step E2: Read the set walking speed parameters and control the walking mechanism to complete the movement of the grinding carriage; Step E3: Based on the distance measurement and pressure values ​​fed back during the walking process, adjust the lateral and longitudinal power mechanisms in real time to achieve feedback control; Step E4: When the distance of the grinding carriage reaches the set value, stop moving and determine whether the current number of grinding passes has reached the set value; if yes, proceed to step E5; if no, accumulate the number of grinding passes, move the grinding carriage in the opposite direction, and proceed to step E3. Step E5: Determine whether to traverse the grinding angles. If yes, end the step; otherwise, proceed to step E1.

9. The method for detecting and grinding repair defects in railway rails according to claim 5, characterized in that: Step 304 achieves pressure stabilization by adjusting the feed depth, as shown in the following formula: Where Δz is the feed longitudinal depth adjustment based on pressure deviation; K p For proportionality coefficient, K i For the integral coefficient, K d are the differential coefficients, the three coefficients are constants, and n is the current sampling time.

10. The method for detecting and grinding repair defects in railway rails according to claim 1, characterized in that: The formula for the Comprehensive Quality Evaluation Index (CGQI) of rail grinding in step 5 is as follows: ,in, , , , For dynamic weights, DPA is the dynamic profile fit; DRC is the defect residue correction degree; SRC is the surface roughness compliance degree; GUI is the grinding quality uniformity score.

Images