Method and model for discriminating steel rail abrasion in real time based on profile alignment

Abstract

The invention discloses a profile alignment-based steel rail abrasion real-time discrimination method and model, and the method comprises the steps: building a steel rail standard profile database model which is used for receiving actual measurement profile data and carrying out the real-time overlapping display; acquiring a point with the minimum vertical coordinate of the outer contour of the steel rail profile from the actually measured profile data, and overlapping the point with the minimum vertical coordinate with the reference point of the standard profile to realize primary alignment of the profile; calculating a perpendicular bisector at the rail waist of the actual measurement profile, and enabling the perpendicular bisector to coincide with the center line of the standard profile to finish rough alignment of the profile; and calculating the center point of the actually measured profile rail web, performing fitting processing to obtain a fitting center line, and enabling the fitting center line to coincide with the standard profile center line to complete profile fine alignment. According to the method, the precision of profile alignment is effectively improved, meanwhile, the influence of profile form errors on an abrasion algorithm is reduced, and then the problem of jumping during alignment of the actually measured profile and the standard profile is solved.

Description

Technical Field

[0001] The present invention relates to the field of track detection technology, and in particular to a real-time rail wear identification method and model based on profile alignment. Background Art

[0002] Rails are core components of railways. The integrity of their profile directly affects the smoothness of train operation, noise, and the risk of derailment. Rail profile refers to the shape of the rail's cross-section. Over long-term service, due to the continuous friction, compression, and impact between the wheels and rails, the rail surface material inevitably wears away, resulting in a gradual change in profile. If this wear is not effectively monitored and controlled, it will significantly increase the risk of train derailment and threaten driving safety. Therefore, how to extend the service life of rails through accurate wear assessment while balancing maintenance costs and operational safety has become a key technical challenge in railway operations and maintenance.

[0003] Traditional inspection methods rely on manual caliper measurements, which are inefficient and lack precision, making it difficult to capture complex profile changes. With the advancement of sensing technology, modern inspection methods such as laser scanning and 3D imaging can quickly acquire high-density point cloud data of the entire rail cross-section, providing a data foundation for quantitative wear analysis. However, achieving high-precision alignment of profile data before and after wear—that is, matching the inspected profile data with the standard profile data—remains crucial for calculating wear. The accuracy of profile alignment directly determines the reliability of wear assessment.

[0004] There are many existing methods for waveform data alignment. Method 1 is a data alignment method based on correlation. This method can achieve horizontal data sampling and supplementation for data with obvious characteristics, thereby achieving consistency in data length. It is not suitable for rail profile alignment. Method 2 is an alignment method based on rail head and rail side feature points. This method will submerge the vertical wear and side wear of the rail and cannot truly reflect the rail wear. Method 3 uses the PCL point cloud registration method to achieve data alignment. This method has good alignment effect, but the efficiency is extremely low and cannot meet real-time requirements. Based on the limitations of the above methods, this paper proposes a fusion alignment method for the rail top, rail side and rail waist data to ensure the stability of profile alignment and accurately calculate the wear amount, thereby guiding maintenance work and ensuring the safety and economy of railway operation. Summary of the Invention

[0005] The purpose of the present invention is to address the current problems of low accuracy and high misjudgment rate in rail wear detection, and to solve the "jump" problem when the measured profile is aligned with the standard profile. This patent invention provides a real-time rail wear identification method and model based on profile alignment, which can filter the measured profile and reduce the impact of profile shape error on rail wear detection.

[0006] The application provides a rail wear real-time identification method based on profile alignment, which is applied to real-time intelligent detection of rail in the field of rail transit. S1, a rail standard profile database model is established, which is used for receiving rail measured profile data for real-time superposition display; S2, a point of minimum longitudinal coordinate of rail profile outer contour is obtained from the measured profile data, and the point of minimum longitudinal coordinate is coincided with the rail top reference point of the standard profile, so that preliminary profile alignment is realized; S3, a center line of the measured profile rail waist is calculated, and the center line is coincided with the center line of the standard profile, so that coarse profile alignment is completed; S4, a center point of the measured profile rail waist is calculated and is subjected to fitting treatment to obtain a fitted center line, and the fitted center line is coincided with the center line of the standard profile, so that fine profile alignment is completed.

[0007] Further, the establishment of the rail standard profile database model comprises the following steps: taking the top center point of the rail standard profile data as a coordinate origin (0, 0), combining the height of the rail top surface and the rail bottom surface, and the standard design data of the rail head, rail waist and rail bottom arc, and constructing the rail standard profile database model.

[0008] Further, after the step S2 and before the step S3, the step further comprises: filtering and smoothing the rail waist data of the measured profile by using a non-recursive filter.

[0009] Further, the step S3 comprises: S31, calculating a minimum distance straight line and a center line of the rail waist of the measured profile; S32, taking the intersection point of the center line and the rail bottom of the measured profile as a reference, the center line is coincided with the center line of the standard profile.

[0010] Further, the step S32 specifically comprises: aligning the intersection horizontal coordinate of the center line and the rail top of the measured profile and the rail bottom of the standard profile with the horizontal coordinate of the rail top reference point of the standard profile, so that the center line is coincided with the center line of the standard profile. Whether the angle of the center line and the center line of the standard profile is zero is confirmed, so that the coincidence effect of the center line and the center line of the standard profile is improved. That is, taking the intersection point of the center line and the rail top of the measured profile as a reference, combining the angle of the center line and the center line of the standard profile, the rail profile is moved, so that the center line is coincided with the center line of the standard profile, and coarse alignment of the rail profile is realized.

[0011] Furthermore, after step S3 and before step S4, the method further includes: processing the left and right rail waists of the measured profile using a vertical coordinate equally spaced interpolation method to make the left and right rail waist data points equal.

[0012] Furthermore, the step S4 includes: S41, calculating all center points corresponding to the left and right rail waists, and performing fitting processing on them to obtain the fitting center line of all rail waist data; S42. Using the intersection of the fitting center line and the measured profile rail top as a reference, rotate the rail profile so that the fitting center line coincides with the standard profile center line.

[0013] Furthermore, the step S42 is specifically as follows: The horizontal coordinate of the intersection of the fitting centerline, the measured rail top, and the standard rail bottom is aligned with the horizontal coordinate of the reference point of the standard rail top, so that the fitting centerline coincides with the standard profile centerline. The coincidence of the fitting centerline and the standard profile centerline is improved by confirming whether the angle between the fitting centerline and the standard profile centerline is zero. Specifically, using the intersection of the fitting centerline and the measured rail top as a reference and combining the angle between the fitting centerline and the standard profile centerline, the rail profile is moved so that the fitting centerline coincides with the standard profile centerline, achieving precise alignment of the rail profile.

[0014] The real-time rail wear identification method based on the profile alignment includes calculating the distance between the measured profile after precise alignment and the standard profile, and identifying the rail wear degree in combination with the rail wear severe damage standard.

[0015] Furthermore, after completing step S4, the measured profile after precise alignment is obtained, and the top surface distance H and the side surface distance L between the measured profile after precise alignment and the standard profile are calculated; The top surface distance H and the side surface distance L are respectively compared with the rail wear severe damage standard, and the rail wear degree is judged according to the difference comparison result. The rail wear degree includes the vertical wear degree and the side wear degree.

[0016] Furthermore, the vertical wear degree includes: When the top surface distance H and the rail wear damage standard deviation value are less than the threshold S H1 When , it is judged that the vertical wear degree is normal; When the top surface distance H and the rail wear damage standard deviation value is greater than or equal to the threshold value S H1 , which is less than the threshold S H2 When the vertical wear degree is judged to be minor; When the top surface distance H and the rail wear damage standard deviation value is greater than or equal to the threshold value S H2When the vertical abrasion degree difference is greater than or equal to the threshold value S

[0017] To improve the rail abrasion discrimination accuracy and reliability, the threshold value S H1 and S H2 may be adaptively set according to the actual application scene in combination with the abrasion injury standard, wherein the threshold value S H1 ranges from 7 to 10, the corresponding S H2 ranges from 8 to 12, the threshold value S H2 is greater than the threshold value S H1 . Taking the 60 rail with a speed less than or equal to 120 km / h as an example, the reference value of S H1 is 9 mm, and the reference value of S H2 is 11 mm; taking the 75 rail with a speed less than or equal to 120 km / h as an example, the reference value of S H1 is 10 mm, and the reference value of S H2 is 12 mm.

[0018] Further, the side surface abrasion degree comprises: When the side surface distance L and the rail abrasion injury standard difference value is less than the threshold value S L1 , the side surface abrasion degree is judged to be normal; When the side surface distance L and the rail abrasion injury standard difference value is greater than or equal to the threshold value S L1 , and less than the threshold value S L2 , the side surface abrasion degree is judged to be slightly injured; When the side surface distance L and the rail abrasion injury standard difference value is greater than or equal to the threshold value S L2 , the side surface abrasion degree is judged to be seriously injured.

[0019] To improve the rail abrasion discrimination accuracy and reliability, the threshold value S L1 and S L2 may be adaptively set according to the actual application scene in combination with the abrasion injury standard, wherein the threshold value S L1 ranges from 9 to 16, the corresponding S L2 ranges from 13 to 21, the threshold value S L2 is greater than the threshold value S L1 . Taking the 60 rail with a speed less than or equal to 120 km / h as an example, the reference value of S L1 is 14 mm, and the reference value of S L2 is 19 mm; taking the 75 rail with a speed less than or equal to 120 km / h as an example, the reference value of S L1 is 16 mm, and the reference value of S L2 is 21 mm.

[0020] Furthermore, the above-mentioned profile alignment method is deployed in a real-time rail wear discrimination model, and the real-time rail wear discrimination model is used to execute the above-mentioned real-time rail wear discrimination method based on profile alignment.

[0021] Furthermore, the rail wear discrimination model receives the measured rail profile data in real time, and uses the above-mentioned real-time rail wear discrimination method based on profile alignment to perform real-time rail wear discrimination to obtain a real-time rail wear degree discrimination result; based on the real-time rail wear degree discrimination result, the rail wear discrimination model is trained and optimized accordingly to obtain an updated rail wear discrimination model.

[0022] Furthermore, the real-time rail wear discrimination model of the target line rail profile data is input into the real-time collection, and the top surface distance H and side distance L of each position are extracted to obtain the top surface distance H set and side distance L set of the rail profile of the line; the top surface distance H set and the side distance L set are respectively subjected to noise reduction processing to filter out interference data; the noise-reduced top surface distance H set and side distance L set are respectively aligned using the above-mentioned profile alignment method, and then the real-time rail wear discrimination method based on profile alignment is used to realize real-time judgment of the wear degree of the rails of the line.

[0023] Compared with the existing technology, the beneficial effects of the present invention are: 1. The present invention makes full use of the effective and stable local features of the full cross-section profile of the rail, performs feature fusion calculations at multiple locations, obtains accurate alignment parameters, and achieves stable alignment effects. 2. The calculation is concise and the reasoning process is fast, which can meet the effect of real-time output of rail wear and make up for the inefficiency of iterative alignment; BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 A schematic flow chart of a rail profile alignment method is provided; Figure 2 The figure is a flow chart of a real-time rail wear identification method based on profile alignment; Figure 3 This is a schematic diagram of the effect of preliminary alignment of rail profiles; Figure 4 This is a schematic diagram of the median vertical line at the waist of the rail profile; Figure 5 This is a schematic diagram of the effect of rough alignment of rail profiles; Figure 6 This is a schematic diagram of the effect of precise alignment of rail profiles. DETAILED DESCRIPTION

[0025] The following provides a clear and complete description of the concept, specific implementation methods and technical effects of the present invention in conjunction with the embodiments and drawings, so as to fully understand the purpose, features and effects of the present invention.

[0026] Example 1 like Figure 1 The present invention discloses a rail profile alignment method, comprising the following steps: S1. Establish a rail standard profile database model for receiving measured rail profile data for real-time overlay display; S2. Obtain the point with the minimum vertical coordinate value of the outer contour of the rail profile from the measured profile data, and overlap the point with the reference point of the top of the standard profile rail to achieve preliminary alignment of the profile. The schematic diagram of the effect of preliminary alignment of the profile is shown in FIG. Figure 3 As shown; S3. Calculate the median perpendicular line at the waist of the measured profile rail, and make the median perpendicular line coincide with the center line of the standard profile to complete the rough alignment of the profile. The effect diagram of the rough alignment of the profile is shown in the figure below. Figure 5 As shown; S4. Calculate the center point of the measured profile rail waist and perform fitting processing to obtain a fitting center line. The fitting center line is overlapped with the standard profile center line to complete the profile fine alignment. The effect diagram of the profile fine alignment is shown in FIG. Figure 6 shown.

[0027] Furthermore, in step S1 of this embodiment, the coordinate system principle is specifically adopted, with the center point of the top surface of the standard rail profile as the coordinate origin (0, 0), and the height of the top surface and the bottom surface of the rail, as well as the standard design data such as the arc of the rail head, rail waist and rail bottom, are combined to construct a standard rail profile database model; the constructed standard rail profile database model is used to receive the real-time collected data of the rail, and the data is superimposed on the standard rail profile database model for display, such as Figure 3 shown.

[0028] Furthermore, after step S2 and before step S3, the method further includes: using a non-recursive filter to perform filtering and smoothing processing on the measured profile rail waist data.

[0029] Furthermore, step S3 includes: S31, calculate the minimum distance straight line and median perpendicular line at the measured rail waist, such as Figure 4 As shown; S32. Using the intersection of the median perpendicular line and the top of the measured profile rail as a reference, the median perpendicular line is made to coincide with the center line of the standard profile.

[0030] Furthermore, the step S32 is specifically as follows: Align the horizontal coordinate of the intersection of the median perpendicular line, the top of the measured rail profile, and the bottom of the standard rail profile with the horizontal coordinate of the reference point of the standard rail profile top, so that the median perpendicular line coincides with the centerline of the standard profile. Verify that the median perpendicular line coincides with the centerline of the standard profile by confirming that the angle between the median perpendicular line and the centerline of the standard profile is zero. That is, using the intersection of the median perpendicular line and the top of the measured rail profile as a reference and combining the angle between the median perpendicular line and the centerline of the standard profile, move the rail profile so that the median perpendicular line coincides with the centerline of the standard profile, achieving a rough alignment of the rail profile.

[0031] Furthermore, after step S3 and before step S4, the method further includes: processing the left and right rail waists of the measured profile using a vertical coordinate equally spaced interpolation method to make the left and right rail waist data points equal.

[0032] Furthermore, the step S4 includes: S41, calculating all center points corresponding to the left and right rail waists, and performing fitting processing on them to obtain the fitting center line of all rail waist data; S42. Using the intersection of the fitting center line and the measured profile rail top as a reference, rotate the rail profile so that the fitting center line coincides with the standard profile center line.

[0033] Furthermore, the step S42 is specifically as follows: The horizontal coordinate of the intersection of the fitting centerline, the measured rail top, and the standard rail bottom is aligned with the horizontal coordinate of the standard rail top reference point, so that the fitting centerline coincides with the standard rail centerline. The alignment of the rail profile with the standard rail profile database model is verified by confirming that the angle between the fitting centerline and the standard rail centerline is zero. Specifically, using the intersection of the fitting centerline and the measured rail top as a reference and the angle between the fitting centerline and the standard rail centerline as a reference, the rail profile is moved so that the fitting centerline coincides with the standard rail centerline, achieving precise alignment of the rail profile. Example

[0034] Based on Example 1, this example proposes a real-time rail wear identification method based on profile alignment, such as Figure 2 As shown in the figure, the real-time intelligent detection of rail wear applied in the field of rail transportation includes the following steps: S1. Establish a rail standard profile database model for receiving measured rail profile data for real-time overlay display; S2. Obtaining the point with the minimum vertical coordinate value of the outer contour of the rail profile from the measured profile data, and aligning the point with the minimum vertical coordinate value with the reference point of the standard profile rail top to achieve preliminary profile alignment; S3, calculate the center line of the measured profile at the rail waist, and coincide the center line with the standard profile center line to complete the rough alignment of the profile; S4, calculate the center point of the measured profile at the rail waist and perform fitting processing to obtain a fitted center line, and coincide the fitted center line with the standard profile center line to complete the fine alignment of the profile; S5, calculate the top surface distance H and the side surface distance L of the measured profile after fine alignment and the standard profile; Based on the rail wear injury standard, the top surface distance H and the side surface distance L are analyzed to determine the degree of rail wear.

[0035] Further, the top surface distance H and the side surface distance L are respectively compared with the rail wear injury standard, and the degree of rail wear is determined according to the difference comparison result, and the degree of rail wear includes the vertical wear degree and the side wear degree.

[0036] Further, the vertical wear degree includes: When the difference between the top surface distance H and the rail wear injury standard is less than a threshold S H1 , it is determined that the vertical wear degree is normal; When the difference between the top surface distance H and the rail wear injury standard is greater than or equal to a threshold S H1 , which is less than a threshold S H2 , it is determined that the vertical wear degree is slightly injured; When the difference between the top surface distance H and the rail wear injury standard is greater than or equal to a threshold S H2 , it is determined that the vertical wear degree is seriously injured.

[0037] To improve the precision and reliability of rail wear determination, the thresholds S H1 and S H2 may be adaptively set according to the actual application scene combined with the wear injury standard, wherein the threshold S H1 ranges from 7 to 10, and the corresponding S H2 ranges from 8 to 12. Taking a 60 rail with a speed less than or equal to 120 km / h as an example, the reference value of S H1 is 9 mm, and the reference value of S H2 is 11 mm; taking a 75 rail with a speed less than or equal to 120 km / h as an example, the reference value of S H1 is 10 mm, and the reference value of S H2 is 12 mm.

[0038] Further, the side wear degree includes: When the difference between the side surface distance L and the rail wear injury standard is less than a threshold S L1 , it is determined that the side wear degree is normal; When the side distance L and the rail wear damage standard deviation are greater than or equal to the threshold S L1 , which is less than the threshold S L2 When the side wear degree is judged to be slightly damaged; When the side distance L and the rail wear damage standard deviation are greater than or equal to the threshold S L2 When the side wear degree is judged to be serious.

[0039] In order to improve the accuracy and reliability of rail wear identification, the threshold S L1 and S L2 It can be set according to the actual application scenario and the wear and tear standard adaptability, where the threshold S L1 The range is [9,16], and the corresponding S L2 The range is [13,21]. Taking the 60-track vehicle with a speed of less than or equal to 120km / h as an example, S L1 The reference value is 14mm, S L2 The reference value is 19mm; taking the 75-gauge with a speed of less than or equal to 120km / h as an example, S L1 The reference value is 16mm, S L2 The reference value is 21mm. Example

[0040] Based on Examples 1 and 2, a real-time rail wear identification model based on profile alignment is constructed. This model is used to implement the aforementioned real-time rail wear identification method based on profile alignment. The infrastructure used to construct the real-time rail wear identification model in this embodiment can utilize open-source models such as DeepSeek and YOLO models, as well as other deep learning models. These are not detailed in this embodiment.

[0041] Furthermore, the rail wear discrimination model receives the measured rail profile data in real time, and uses the above-mentioned real-time rail wear discrimination method based on profile alignment to perform real-time rail wear discrimination to obtain a real-time rail wear degree discrimination result; based on the real-time rail wear degree discrimination result, the rail wear discrimination model is trained and optimized accordingly to obtain an updated rail wear discrimination model.

[0042] Furthermore, the real-time rail wear discrimination model of the target line rail profile data is input into the real-time collection, and the top surface distance H and side distance L of each position are extracted to obtain the top surface distance H set and side distance L set of the rail profile of the line; the top surface distance H set and the side distance L set are respectively subjected to noise reduction processing to filter out interference data; the noise-reduced top surface distance H set and side distance L set are respectively aligned using the above-mentioned profile alignment method, and then the real-time rail wear discrimination method based on profile alignment is used to realize real-time judgment of the wear degree of the rails of the line.

[0043] Compared with the existing technology, the beneficial effects of the present invention are: 1. The present invention makes full use of the effective and stable local features of the full cross-section profile of the rail, performs feature fusion calculations at multiple locations, obtains accurate alignment parameters, and achieves stable alignment effects. 2. The calculation is concise and the reasoning process is fast, which can meet the effect of real-time output of rail wear and make up for the inefficiency of iterative alignment; The above-described embodiments merely represent specific implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of protection of the present application. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the technical concept of the present application, and all such variations and improvements fall within the scope of protection of the present application.

Claims

1. A real-time rail wear identification method based on profile alignment, characterized in that: The steps include: S1. Establish a rail standard profile database model for receiving measured rail profile data for real-time overlay display; S2. Obtaining the point with the minimum vertical coordinate value of the outer contour of the rail profile from the measured profile data, and aligning the point with the minimum vertical coordinate value of the outer contour of the rail profile with the reference point of the standard profile rail top to achieve preliminary profile alignment; S3, calculating the median perpendicular line of the measured profile rail waist, and aligning the median perpendicular line with the center line of the standard profile to achieve rough alignment of the profile; S4. Calculate the center point of the measured rail waist and perform fitting processing to obtain a fitting center line, and overlap the fitting center line with the standard profile center line to achieve precise profile alignment; S5. Calculate the distance between the measured profile after precise alignment and the standard profile, and determine the degree of rail wear based on the rail wear severity standard.

2. The method for real-time rail wear identification based on profile alignment according to claim 1, characterized in that: The step S3 comprises: S31, calculating the minimum distance straight line and the perpendicular midline at the measured rail waist; S32. Using the intersection of the median perpendicular line and the top of the measured profile rail as a reference, the median perpendicular line is made to coincide with the center line of the standard profile.

3. The method for real-time rail wear identification based on profile alignment according to claim 1, characterized in that: After step S2 and before step S3, the method further includes: The non-recursive filter is used to filter and smooth the measured rail waist data.

4. The method for real-time rail wear identification based on profile alignment according to claim 1, characterized in that: The step S4 comprises: S41, calculating all center points corresponding to the left and right rail waists, and performing fitting processing on them to obtain the fitting center line of all rail waist data; S42. Using the intersection of the fitting center line and the measured profile rail top as a reference, rotate the rail profile so that the fitting center line coincides with the standard profile center line.

5. The method for real-time rail wear identification based on profile alignment according to claim 1, characterized in that: After step S3 and before step S4, the method further includes: The vertical coordinate equal-interval interpolation method is used to process the left and right rail waists of the measured profile so that the data points on the left and right rail waists are equal.

6. The method for real-time rail wear identification based on profile alignment according to claim 1, characterized in that: The S5 specifically includes: Calculate the top surface distance H and side surface distance L between the measured profile after precise alignment and the standard profile; compare the top surface distance H and side surface distance L with the rail wear severe damage standard, and judge the rail wear degree based on the difference comparison result, wherein the rail wear degree includes the vertical wear degree and the side wear degree.

7. The method for real-time rail wear identification based on profile alignment according to claim 6, characterized in that: The vertical wear degree includes: When the top surface distance H and the rail wear damage standard deviation value are less than the threshold S H1 When , it is judged that the vertical wear degree is normal; When the top surface distance H and the rail wear damage standard deviation value is greater than or equal to the threshold value S H1 , which is less than the threshold S H2 When the vertical wear degree is judged to be minor; When the top surface distance H and the rail wear damage standard deviation value is greater than or equal to the threshold value S H2 When the vertical wear degree is judged to be serious injury.

8. The method for real-time rail wear identification based on profile alignment according to claim 6, characterized in that: The side wear degree includes: When the side distance L and the rail wear damage standard deviation are less than the threshold S L1 When the wear degree of the side surface is judged to be normal; When the side distance L and the rail wear damage standard deviation are greater than or equal to the threshold S L1 , which is less than the threshold S L2 When the side wear degree is judged to be slightly damaged; When the side distance L and the rail wear damage standard deviation are greater than or equal to the threshold S L2 When the side wear degree is judged to be serious.

9. A real-time rail wear identification model based on profile alignment, characterized in that: The rail wear real-time discrimination model is used to execute the rail wear real-time discrimination method based on profile alignment as described in claims 1 to 8.

10. The real-time rail wear identification model based on profile alignment according to claim 9, characterized in that: The rail wear discrimination model receives the measured rail profile data in real time, and uses the real-time rail wear discrimination method based on profile alignment as described in any one of claims 1 to 8 to perform real-time rail wear discrimination to obtain a real-time rail wear degree discrimination result; Based on the real-time rail wear degree judgment result, the rail wear judgment model is trained and optimized accordingly to obtain the updated rail wear judgment model.

Images