A high-speed rail pantograph carbon slide plate abrasion detection method

Abstract

A high-speed rail pantograph carbon slide plate abrasion detection method, comprising: acquiring high-speed rail pantograph carbon slide plate point cloud profile data through a single laser sensor; processing the data using a motion blur correction algorithm based on spatial range screening and reflection intensity screening to obtain effective point cloud data; statistically filtering the effective point cloud data, then performing normalization processing to obtain normalized effective point cloud data; using reflection characteristic identification method and feature point identification method respectively to obtain the first and second demarcation points of the pantograph horn and the carbon slide plate; screening and integrating to obtain the final demarcation point, and then intercepting to obtain the measured carbon slide plate point cloud data set; using the point-to-plane ICP registration method to compare and register the measured carbon slide plate point cloud data set with the standard carbon slide plate point cloud data set, so as to calculate the abrasion value and the over-limit abrasion area of the pantograph carbon slide plate. The present application can realize efficient and high-precision detection of high-speed rail pantograph carbon slide plate abrasion.

Description

Technical Field

[0001] This invention relates to the field of rail transit testing technology, specifically to a method for detecting wear of carbon sliding plates on high-speed rail pantographs. Background Technology

[0002] With the rapid development of high-speed rail, the reliability and safety of high-speed rail operation are of paramount importance. As a crucial component of the power supply system during high-speed rail operation, the wear and tear of the pantograph's carbon contactor directly affects the stability and safety of power acquisition. Severe wear of the carbon contactor can easily lead to arcing and disconnection, and even cause major traffic accidents. Therefore, how to effectively and accurately detect the wear of the carbon contactor of the high-speed rail pantograph is a critical issue.

[0003] Traditional methods for detecting wear on the carbon contact plate of a pantograph are mainly manual, i.e., contact-based detection. The basic steps include driving the locomotive into the depot, stopping the locomotive, lowering the pantograph, and disconnecting the power. Then, workers climb onto the roof of the locomotive and use professional instruments to measure the carbon contact plate. Although this method is simple and convenient, it is relatively inefficient, labor-intensive, and cannot achieve real-time monitoring. In addition, the test results are affected by the subjective judgment of the staff.

[0004] To address the numerous problems and shortcomings of manual inspection, and to meet the needs of automated inspection in rail transit, non-contact inspection has been applied to pantograph wear detection. Currently, non-contact inspection methods mainly include ultrasonic testing, three-dimensional binocular vision testing, and image detection. Ultrasonic testing is easily affected by ambient noise, and three-dimensional binocular vision testing is currently not mature enough and has limited application. Image detection is widely used, but it is affected by lighting conditions and the measurements are intermittent and discontinuous, leading to missed detections. Therefore, how to achieve efficient and high-precision detection of carbon contact plate wear on high-speed railway pantographs is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method for detecting the wear of carbon sliding plates of pantographs in high-speed railways, so as to achieve efficient and high-precision detection of the wear of carbon sliding plates of pantographs in high-speed railways.

[0006] A method for detecting wear of carbon sliding plates in high-speed rail pantographs includes:

[0007] Step S1: Acquire point cloud profile data of the carbon sliding plate of the high-speed rail pantograph in PCD format using a single-line laser sensor;

[0008] Step S2: Spatial range and reflection intensity are filtered for the carbon skateboard point cloud profile data. Then, the motion fuzz correction algorithm is used to process the filtered data to obtain effective point cloud data.

[0009] Step S3: Perform statistical filtering on the effective point cloud data, and then normalize it to obtain normalized effective point cloud data.

[0010] Step S4: Based on the normalized effective point cloud data, the reflection characteristic identification method is used to obtain the first dividing point between the pantograph horn and the carbon sliding plate.

[0011] Step S5: Based on the normalized effective point cloud data, the feature point recognition method is used to obtain the second dividing point between the pantograph horn and the carbon sliding plate.

[0012] Step S6: The first and second boundary points are screened and integrated by combining confidence and geometric consistency tests to obtain the final boundary point between the pantograph horn and the carbon skateboard. Based on the final boundary point, the measured carbon skateboard point cloud dataset is obtained.

[0013] Step S7: The measured carbon skateboard point cloud dataset is compared and registered with the standard carbon skateboard point cloud dataset using the point-to-surface ICP registration method. Then, Lie algebra is used for rotation, and Gauss-Newton framework is used for iterative solution to calculate the wear value and over-limit wear area of ​​the pantograph carbon skateboard.

[0014] The method for detecting wear of carbon sliding plates in high-speed rail pantographs provided by the present invention has the following beneficial effects:

[0015] (1) Since laser sensors are usually in constant motion, the acquisition time and pose of points in the same frame are different, which will cause motion blur. In order to eliminate its influence, this invention uses a motion blur correction algorithm based on spatial range filtering and reflection intensity filtering to process the carbon skateboard point cloud profile data to obtain effective point cloud data. Under the condition of known motion kernel, the true height can be restored in one go. It has real-time performance, numerical stability and edge fidelity, and provides reliable reference data for subsequent high-precision boundary and wear calculation.

[0016] (2) The present invention uses the reflection characteristic recognition method and the feature point recognition method to obtain the boundary point between the pantograph horn and the carbon skateboard, and uses the confidence level and geometric consistency test combined to screen and integrate them, so as to obtain the final boundary point with higher confidence. Then, based on the final boundary point, the measured carbon skateboard point cloud dataset is obtained, which can ensure the validity and accuracy of the measured carbon skateboard point cloud dataset.

[0017] (3) The present invention uses the point-to-surface ICP registration method to compare and register the measured carbon sliding plate point cloud dataset with the standard carbon sliding plate point cloud dataset, then uses Lie algebra for rotation, and uses the Gauss-Newton framework for iterative solution, which can more accurately calculate the wear value of the pantograph carbon sliding plate.

[0018] (4) The present invention uses a single-line laser sensor, which has the advantages of being resistant to sunlight, rain and snow and requiring no supplemental light at night. At the same time, it has high data accuracy and high sampling frame rate. The present invention is a non-contact detection method, which has higher detection efficiency and lower labor costs. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the method for detecting the wear of the carbon sliding plate of a high-speed rail pantograph provided in an embodiment of the present invention. Detailed Implementation

[0020] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain embodiments of the present invention, and should not be construed as limiting the present invention.

[0021] Please see Figure 1 The present invention provides a method for detecting wear of carbon sliding plates of high-speed rail pantographs, comprising steps S1 to S7:

[0022] Step S1: Acquire point cloud profile data of the carbon sliding plate of the high-speed rail pantograph in PCD format using a single-line laser sensor.

[0023] In this embodiment, two single-line laser sensors are used to detect the point cloud profile data of the carbon sliding plate of the high-speed rail pantograph. This achieves complementary occlusion and data redundancy without increasing the scanning cycle, improving the integrity of the pantograph's point cloud profile and enhancing lateral and vertical repeatability. Specifically, one single-line laser sensor is installed on each side of the overhead contact line support gantry, with the sensor axis perpendicular to the track direction. A point laser sensor is installed in front of the single-line laser sensors, directly facing the oncoming train, to detect pantograph approach signals. All sensors support three-degree-of-freedom fine-tuning to ensure the laser beam accurately covers the pantograph's carbon sliding plate area. The single-line laser sensors have a wavelength of 650nm, a sampling frequency ≥50kHz, and their axes are perpendicular to the track direction.

[0024] When the high-speed train enters the detection area, the point laser sensor captures the pantograph's approach signal and outputs a high-level pulse to the single-line laser sensor, triggering it to begin scanning. The point laser sensor has a trigger distance of 10m and a response time of ≤1ms. The single-line laser sensor continuously scans the surface of the pantograph's carbon sliding plate at a density of 5000 points / frame, generating PCD format point cloud profile data of the high-speed train's pantograph's carbon sliding plate, including three-dimensional coordinates and timestamps.

[0025] Step S2 involves filtering the spatial range and reflection intensity of the carbon skateboard point cloud profile data, and then using a motion fuzz correction algorithm to process the filtered data to obtain valid point cloud data.

[0026] Specifically, data within the valid range is obtained by filtering by spatial range, and data that meets the reflection intensity threshold is obtained by filtering by reflection intensity.

[0027] In addition, during the data acquisition process, the lidar is usually in constant motion, and the acquisition time and pose of points in the same frame are also different, resulting in motion distortion in the point cloud data. Therefore, motion blur correction is required for the selected data.

[0028] In this embodiment, during the processing of the filtered data using the motion blur correction algorithm, the expression for discrete convolution is:

[0029]

[0030] in, In the PCD format of the carbon sliding plate point cloud profile data of the high-speed rail pantograph The fuzzy height at that location, This represents the discrete depth index. , The total number of ranging points that meet the reflection intensity threshold range and whose spatial height is within the effective interval, For scan resolution, This indicates that the fuzzy kernel is used in the convolution operation. Each weight corresponds to a depth The actual height sampled value at that location, This is the offset. ; This represents the total number of sampling points taken when discretizing the motion blur kernel, and determines the width of the blur window; For uniform weighting, , These are discrete noise samples.

[0031] Step S3: Perform statistical filtering on the effective point cloud data, and then normalize it to obtain normalized effective point cloud data.

[0032] The process involves first performing statistical filtering on the effective point cloud data to remove noise and smooth the data, and then performing normalization to obtain the normalized effective point cloud data.

[0033] Step S4: Based on the normalized effective point cloud data, the reflection characteristic identification method is used to obtain the first dividing point between the pantograph horn and the carbon sliding plate.

[0034] Since there is a significant difference in the reflection intensity between the carbon sliding plate and the pantograph horn, the first dividing point between the pantograph horn and the carbon sliding plate can be obtained based on the material reflection characteristics and the reflection characteristic identification method.

[0035] Step S4 specifically includes:

[0036] Laser scanning time-series sorting was performed on the normalized point cloud data to construct a reflection intensity sequence. , elements in Indicates the first Normalized reflection intensity values ​​of each effective ranging point. Satisfy the following formula:

[0037]

[0038]

[0039] in, This represents the normalized reflection intensity after distance-squared compensation of the original echo intensity. For single-line laser sensors in the first The original reflection intensity values ​​actually measured at each effective point. For reference distance, For single-line laser sensors to the first Euclidean distance to each valid point;

[0040] To address the uneven point cloud density, interpolation is required. In this embodiment, the reflection intensity sequence is used. As the input source, first, at fixed time intervals on the time axis. Resampling is performed to obtain an equidistant time interval sequence. A sliding window is then introduced onto this sequence, and the local standard deviation and gradient jump are calculated point-by-point. A candidate point set is selected based on the local standard deviation and gradient jump, where the sampled points in the candidate point set satisfy the following equation:

[0041]

[0042]

[0043]

[0044]

[0045] in, The first time interval sequence after resampling Local standard deviation of each sampling point The first time interval sequence after resampling Gradient jump at each sampling point The mean of the global standard deviation. The global gradient mean. and In this embodiment, the coefficient is an empirical factor. Take 2.0, Take 1.5, This represents the total number of sampling points in the resampled, equally spaced sequence.

[0046] To obtain more accurate points, the spatial location of the data is verified in the candidate point set to eliminate false peaks caused by multiple adjacent points exceeding the threshold simultaneously. Finally, the first boundary point between the pantograph horn and the carbon sliding plate is obtained, which satisfies the following formula:

[0047]

[0048]

[0049] in, This is the first dividing point. This represents the minimum width of the carbon skateboard. This represents the maximum width of the carbon skateboard. The first dividing point The derivative of height with respect to width at a given point. This is the lower limit threshold for the rate of change used to eliminate spurious peaks.

[0050] Step S5: Based on the normalized effective point cloud data, the feature point recognition method is used to obtain the second dividing point between the pantograph horn and the carbon sliding plate.

[0051] Among them, based on the changes in geometric features, the feature point recognition method is used to obtain the second dividing point between the pantograph horn and the carbon sliding plate.

[0052] Step S5 specifically includes:

[0053] For any point in the normalized valid point cloud data Calculation points The average coordinates of all points in the neighborhood , will point coordinates minus Obtaining deviation According to the deviation Calculation points The covariance is obtained, and then the covariance matrix is ​​obtained. The covariance matrix is ​​decomposed into eigenvalues ​​to solve for the eigenvalues ​​and eigenvectors, so as to obtain the minimum eigenvalue and the corresponding eigenvector.

[0054] Take the eigenvector corresponding to the smallest eigenvalue as the normal vector, and the two eigenvectors orthogonal to the normal vector. , To form a tangential base, , With u and v axes as the u and v axes and z as the normal vector, a local right-handed coordinate system is established. The coordinates of the neighborhood points are projected onto the local right-handed coordinate system to obtain the local two-dimensional coordinates and relative elevations. A quadratic surface is fitted in the local right-handed coordinate system using the least squares method. The surface parameters are obtained by solving the least squares solution. The first derivative of the surface is then calculated using the surface parameters. Based on the first derivative of the surface, a shape operator matrix is ​​constructed.

[0055] After eigendecomposition of the shape operator matrix, the principal curvature and principal direction are obtained, and then the mean absolute curvature is derived.

[0056] The average absolute curvature of each point is rearranged into a one-dimensional curvature signal according to the scanning order. Based on the one-dimensional curvature signal, a first-order difference signal is obtained, expressed as:

[0057]

[0058] in, For point The corresponding one-dimensional curvature signal, For point The corresponding one-dimensional curvature signal, For point The corresponding first-order differential signal;

[0059] The first-order difference signal is subjected to noise suppression using the Gaussian smoothing method to obtain a smoothed difference signal, expressed as follows:

[0060]

[0061] in, For point The corresponding smoothed differential signal, For a standard deviation of Gaussian filter;

[0062] Peak index set is selected based on the smoothed difference signal. Among them, the peak index set The smoothed differential signal in the equation satisfies the following formula:

[0063]

[0064]

[0065]

[0066]

[0067] in, For the minimum peak spacing, For point The corresponding smoothed differential signal, For point The corresponding smoothed differential signal, For signal threshold, The mean of the signal. It is a constant. The standard deviation of the signal;

[0068] The index of the second boundary point is obtained based on the peak index set. The expression is:

[0069]

[0070] in, This indicates the position where the maximum value is taken;

[0071] Index based on the second boundary point The coordinates of the second dividing point between the pantograph horn and the carbon slide plate are obtained.

[0072] Step S6: The first and second boundary points are screened and integrated using a combination of confidence level and geometric consistency test to obtain the final boundary point between the pantograph horn and the carbon skateboard. Based on the final boundary point, the measured carbon skateboard point cloud dataset is obtained.

[0073] Specifically, step S6 includes:

[0074] Comparing the first and second boundary points, if the spatial distance between the first and second boundary points is less than a preset distance threshold, the average coordinates of the first and second boundary points are taken as the coordinates of the final boundary point; if the spatial distance between the first and second boundary points is not less than the preset distance threshold, the one with higher confidence is selected as the final boundary point.

[0075] Based on the final dividing point, the measured carbon skateboard point cloud dataset was obtained.

[0076] Step S7: The measured carbon skateboard point cloud dataset is compared and registered with the standard carbon skateboard point cloud dataset using the point-to-surface ICP registration method. Then, Lie algebra is used for rotation, and Gauss-Newton framework is used for iterative solution to calculate the wear value and over-limit wear area of ​​the pantograph carbon skateboard.

[0077] Specifically, step S7 includes:

[0078] Define the measured carbon skateboard point cloud dataset as follows: The standard carbon skateboard point cloud dataset is Then, a nonlinear least squares objective function is constructed, expressed as:

[0079]

[0080] in, This indicates taking the minimum value. This is the index of the point cloud dataset of the measured carbon skateboard. This refers to the index of the standard carbon skateboard point cloud dataset. The number of measured profile points is used to enter the point-to-surface ICP registration process. These are the weighting coefficients. for The first in One point, for Neutral The point with the closest Euclidean distance. For point The unit normal vector at that location, Indicates transpose. Let be a rotation matrix. It is a translation vector;

[0081] Based on the nonlinear least squares objective function, Lie algebra is used. The rotation matrix to be optimized Parameterization is performed to obtain the nonlinear residual function. A first-order Taylor expansion is then performed on the nonlinear residual function, and the state increment, residual vector, Jacobian matrix, and normal equation are calculated sequentially. Finally, the state increment is iteratively solved using the Gauss-Newton framework, and the results are updated. and until convergence;

[0082] After the iteration converges, the iteration stops, and the final optimal rotation matrix is ​​obtained. and optimal translation vector This allows us to obtain the wear value of the pantograph carbon slide plate along the normal direction. This process eliminates tangential errors and reuses the corresponding relationships, integrating wear value calculation and registration, and improving accuracy. The specific expression is:

[0083]

[0084] Select items whose wear value along the normal direction is greater than the preset maximum wear depth threshold. The area in question is the area of ​​excessive wear.

[0085] In summary, the high-speed rail pantograph carbon sliding plate wear detection method according to the above embodiments has the following beneficial effects:

[0086] (1) Since laser sensors are usually in constant motion, the acquisition time and pose of points in the same frame are different, which will cause motion blur. In order to eliminate its influence, this invention uses a motion blur correction algorithm based on spatial range filtering and reflection intensity filtering to process the carbon skateboard point cloud profile data to obtain effective point cloud data. Under the condition of known motion kernel, the true height can be restored in one go. It has real-time performance, numerical stability and edge fidelity, and provides reliable reference data for subsequent high-precision boundary and wear calculation.

[0087] (2) The present invention uses the reflection characteristic recognition method and the feature point recognition method to obtain the boundary point between the pantograph horn and the carbon skateboard, and uses the confidence level and geometric consistency test combined to screen and integrate them, so as to obtain the final boundary point with higher confidence. Then, based on the final boundary point, the measured carbon skateboard point cloud dataset is obtained, which can ensure the validity and accuracy of the measured carbon skateboard point cloud dataset.

[0088] (3) The present invention uses the point-to-surface ICP registration method to compare and register the measured carbon sliding plate point cloud dataset with the standard carbon sliding plate point cloud dataset, then uses Lie algebra for rotation, and uses the Gauss-Newton framework for iterative solution, which can more accurately calculate the wear value of the pantograph carbon sliding plate.

[0089] (4) The present invention uses a single-line laser sensor, which has the advantages of being resistant to sunlight, rain and snow and requiring no supplemental light at night. At the same time, it has high data accuracy and high sampling frame rate. The present invention is a non-contact detection method, which has higher detection efficiency and lower labor costs.

[0090] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A method for detecting wear of carbon sliding plates in high-speed rail pantographs, characterized in that, include: Step S1: Acquire point cloud profile data of the carbon sliding plate of the high-speed rail pantograph in PCD format using a single-line laser sensor; Step S2: Spatial range and reflection intensity are filtered for the carbon skateboard point cloud profile data. Then, the motion fuzz correction algorithm is used to process the filtered data to obtain effective point cloud data. Step S3: Perform statistical filtering on the effective point cloud data, and then normalize it to obtain normalized effective point cloud data. Step S4: Based on the normalized effective point cloud data, the reflection characteristic identification method is used to obtain the first dividing point between the pantograph horn and the carbon sliding plate. Step S5: Based on the normalized effective point cloud data, the feature point recognition method is used to obtain the second dividing point between the pantograph horn and the carbon sliding plate. Step S6: The first and second boundary points are screened and integrated by combining confidence and geometric consistency tests to obtain the final boundary point between the pantograph horn and the carbon skateboard. Based on the final boundary point, the measured carbon skateboard point cloud dataset is obtained. Step S7: The measured carbon skateboard point cloud dataset is compared and registered with the standard carbon skateboard point cloud dataset using the point-to-surface ICP registration method. Then, Lie algebra is used for rotation, and Gauss-Newton framework is used for iterative solution to calculate the wear value and over-limit wear area of ​​the pantograph carbon skateboard. Specifically, step S4 includes: Laser scanning time-series sorting was performed on the normalized point cloud data to construct a reflection intensity sequence. , elements in Indicates the first Normalized reflection intensity values ​​of each effective ranging point; Then, using the reflection intensity sequence As the input source, first, at fixed time intervals on the time axis. Resampling is performed to obtain an equal-time interval sequence. Then, a sliding window is introduced on the equal-time interval sequence to calculate the local standard deviation and gradient jump point by point. Candidate point set is selected based on the local standard deviation and gradient jump. The spatial location of the data is verified in the candidate point set to eliminate false peaks caused by multiple adjacent points exceeding the threshold at the same time, and finally the first dividing point between the pantograph horn and the carbon slide plate is obtained. Step S5 specifically includes: For any point in the normalized valid point cloud data Calculation points The average coordinates of all points in the neighborhood , will point coordinates minus Obtaining deviation According to the deviation Calculation points The covariance is obtained, and then the covariance matrix is ​​obtained. The covariance matrix is ​​decomposed into eigenvalues ​​to solve for the eigenvalues ​​and eigenvectors, so as to obtain the minimum eigenvalue and the corresponding eigenvector. Take the eigenvector corresponding to the smallest eigenvalue as the normal vector, and the two eigenvectors orthogonal to the normal vector. , To form a tangential base, , With u and v axes as the u and v axes and z as the normal vector, a local right-handed coordinate system is established. The coordinates of the neighborhood points are projected onto the local right-handed coordinate system to obtain the local two-dimensional coordinates and relative elevations. A quadratic surface is fitted in the local right-handed coordinate system using the least squares method. The surface parameters are obtained by solving the least squares solution. The first derivative of the surface is then calculated using the surface parameters. Based on the first derivative of the surface, a shape operator matrix is ​​constructed. After eigendecomposition of the shape operator matrix, the principal curvature and principal direction are obtained, and then the mean absolute curvature is derived. The average absolute curvature of each point is rearranged into a one-dimensional curvature signal according to the scanning order, and a first-order difference signal is obtained based on the one-dimensional curvature signal. The first-order difference signal is subjected to noise suppression using the Gaussian smoothing method to obtain a smoothed difference signal; Peak index set is selected based on the smoothed difference signal. ; The index of the second boundary point is obtained based on the peak index set. ; Index based on the second boundary point The coordinates of the second dividing point between the pantograph horn and the carbon slide plate are obtained.

2. The method for detecting wear of carbon sliding plates in high-speed rail pantographs according to claim 1, characterized in that, In step S2, during the processing of the selected data using the motion blur correction algorithm, the expression for discrete convolution is: in, In the PCD format of the carbon sliding plate point cloud profile data of the high-speed rail pantograph The fuzzy height at that point, This represents the discrete depth index. , The total number of ranging points that meet the reflection intensity threshold range and whose spatial height is within the effective interval, For scan resolution, This indicates that the fuzzy kernel is used in the convolution operation. Each weight corresponds to a depth The actual height sampled value at that location, This is the offset. , This represents the total number of sampling points taken when discretizing the motion fuzzy kernel. For uniform weighting, , These are discrete noise samples.

3. The method for detecting wear of carbon sliding plates in high-speed rail pantographs according to claim 2, characterized in that, In step S4, Satisfy the following formula: in, This represents the normalized reflection intensity after distance-squared compensation of the original echo intensity. For single-line laser sensors in the first The original reflection intensity values ​​actually measured at each effective point. For reference distance, For single-line laser sensors to the first Euclidean distance to each valid point; The sampling points in the candidate point set satisfy the following formula: in, The first time interval sequence after resampling Local standard deviation of each sampling point The first time interval sequence after resampling Gradient jump at each sampling point The mean of the global standard deviation. The global gradient mean. and This is an empirical coefficient. This represents the total number of sampling points in the resampled, equally spaced sequence. The first dividing point satisfies the following equation: in, This is the first dividing point. This represents the minimum width of the carbon skateboard. This represents the maximum width of the carbon skateboard. The first dividing point The derivative of height with respect to width at a given point. This is the lower limit threshold for the rate of change used to eliminate spurious peaks.

4. The method for detecting wear of carbon sliding plates in high-speed rail pantographs according to claim 3, characterized in that, In step S5, a first-order difference signal is obtained based on the one-dimensional curvature signal, and its expression is: in, For point The corresponding one-dimensional curvature signal, For point The corresponding one-dimensional curvature signal, For point The corresponding first-order differential signal; The first-order difference signal is subjected to noise suppression using the Gaussian smoothing method to obtain a smoothed difference signal, expressed as follows: in, For point The corresponding smoothed differential signal, For a standard deviation of Gaussian filter; Peak index set is selected based on the smoothed difference signal. Among them, the peak index set The smoothed differential signal in the equation satisfies the following formula: in, For the minimum peak spacing, For point The corresponding smoothed differential signal, For point The corresponding smoothed differential signal, For signal threshold, The mean of the signal. It is a constant. The standard deviation of the signal; The index of the second boundary point is obtained based on the peak index set. The expression is: in, This indicates the position where the maximum value is taken.

5. The method for detecting wear of carbon sliding plates in high-speed rail pantographs according to claim 4, characterized in that, Step S6 specifically includes: Comparing the first and second boundary points, if the spatial distance between the first and second boundary points is less than a preset distance threshold, the average coordinates of the first and second boundary points are taken as the coordinates of the final boundary point; if the spatial distance between the first and second boundary points is not less than the preset distance threshold, the one with higher confidence is selected as the final boundary point. Based on the final dividing point, the measured carbon skateboard point cloud dataset was obtained.

6. The method for detecting wear of carbon sliding plates in high-speed rail pantographs according to claim 5, characterized in that, Step S7 specifically includes: Define the measured carbon skateboard point cloud dataset as follows: The standard carbon skateboard point cloud dataset is Then, a nonlinear least squares objective function is constructed, expressed as: in, This indicates taking the minimum value. This is the index of the point cloud dataset of the measured carbon skateboard. This refers to the index of the standard carbon skateboard point cloud dataset. The number of measured profile points is used to enter the point-to-surface ICP registration process. These are the weighting coefficients. for The first in One point, for Neutral The point with the closest Euclidean distance. For point The unit normal vector at that location, Indicates transpose. Let be a rotation matrix. It is a translation vector; Based on the nonlinear least squares objective function, Lie algebra is used. The rotation matrix to be optimized Parameterization is performed to obtain the nonlinear residual function. A first-order Taylor expansion is then performed on the nonlinear residual function, and the state increment, residual vector, Jacobian matrix, and normal equation are calculated sequentially. Finally, the state increment is iteratively solved using the Gauss-Newton framework, and the results are updated. and until convergence; After the iteration converges, the iteration stops, and the final optimal rotation matrix is ​​obtained. and optimal translation vector This allows us to obtain the wear value of the pantograph carbon slide plate along the normal direction. The expression is: Select items whose wear value along the normal direction is greater than the preset maximum wear depth threshold. The area in question is the area of ​​excessive wear.

Images