Ultrasonic phased array full focus mechanical arm scanning method for detecting defects in a welded joint of a seamless rail

Abstract

The application provides an ultrasonic phased array full-focus mechanical arm scanning method for seamless rail weld defects, and relates to the field of weld detection, and comprises the following steps: S1, the flaw detection vehicle travels on the rail route, and a laser scanner synchronously scans the rail, and the flaw detection vehicle stops when the rail weld reinforcement is collected; S2, an upper computer corrects a preset scanning path point set by using an automatic guiding positioning algorithm to obtain a new scanning path point set to generate a maintenance path; S3, the upper computer drives a mechanical arm to clamp an ultrasonic phased array probe to scan along the maintenance path; and S4, the ultrasonic phased array displays a real-time full-focus imaging diagram according to the signal data collected by scanning and can record the detected defects. The application realizes the automatic scanning of the full-section position of the rail weld defects, and uses the automatic guiding positioning technology to correct the scanning path of the mechanical arm, so that the positioning error influence of the scanning program of the mechanical arm in the automatic application is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rail weld detection, in particular to an ultrasonic phased array full-focus mechanical arm scanning method for detecting defects of seamless rail welds. BACKGROUND

[0002] The track directly bears various loads from the locomotive vehicle and is the basis for train operation. The track bears the repeated action of the dynamic load of the locomotive vehicle. In this process, the rail weld, especially in the seamless line, may produce various damages. If the damages are not detected and repaired in time, the rail weld breakage accident is likely to occur. In recent years, nearly two-thirds of the rail breakage accidents occur at the rail weld and the heat affected zone. The detection of the rail weld has attracted great attention from the railway supervision department. Considering the short sky window time and the need for full coverage scanning of the seamless rail weld, and the difficulty in detecting the orientation defects represented by the vertical crack defects.

[0003] The existing technical solutions for non-destructive testing of seamless rails mainly include large-scale rail detection vehicles for regular inspection, electric detection trolleys, hand-push detection vehicles, and manual inspection methods. The rail detection vehicles and electric detection trolleys have high speed, but they cannot comprehensively detect the weld defects and have detection blind spots. The hand-push detection vehicles have low detection speed. Since the arrangement position of the probe is fixed, the traditional detection combined probe scheme is adopted, and there is a blind area in the detection of the rail bottom part. At present, only manual holding of the probe can achieve comprehensive defect detection of the rail weld. However, the detection rate of the rail gullet excess height defects and the rail bottom defects is low, and the detection efficiency of the detection personnel working at night is low and easy to fatigue. The detection results are affected by the level, attitude, and fatigue degree of the detection personnel. The rail detection vehicles and electric detection trolleys have high speed, but they cannot comprehensively detect the weld defects and have detection blind spots. The hand-push detection vehicles have low detection speed. Since the arrangement position of the probe is fixed, the traditional detection combined probe scheme is adopted, and there is a blind area in the detection of the rail bottom part. At present, only manual holding of the probe can achieve comprehensive defect detection of the rail weld. However, the detection rate of the rail gullet excess height defects and the rail bottom defects is low, and the detection efficiency of the detection personnel working at night is low and easy to fatigue. The detection results are affected by the level, attitude, and fatigue degree of the detection personnel.

[0004] The ultrasonic phased array full focusing technology is an advanced ultrasonic detection technology integrating high precision and high resolution. The technology records all the ultrasonic pulses received by each crystal of the probe corresponding to the emission of each crystal, and constructs high-resolution imaging according to the time delay of each point in the imaging area. As a high-level method in the field of ultrasonic imaging, the full focusing technology has the unique feature of using all the crystals of the full aperture of the probe to generate full matrix capture through synthetic beam forming, and the use of delay ensures the focused imaging of each point in the area, thereby constructing an image with significantly improved clarity. This technology not only greatly enhances the resolution and signal-to-noise ratio of the image, but also makes it possible to detect small defects. The traditional phased array detection has good imaging effect for defects at the focusing depth, but the imaging of defects at other positions will be distorted and the amplitude will be reduced, which is not conducive to defect quantification and small defect detection. The phased array full focusing imaging technology realizes high focusing in the specified detection area, which is conducive to better simultaneous detection of a larger range.

[0005] Chinese Patent Application Publication No. CN116465965A discloses a steel rail defect detection method based on phased array ultrasonic, which includes: a probe for detecting a steel rail is a 128-element phased array ultrasonic probe, which is installed on a mechanical arm that can move along the longitudinal direction of the steel rail. The elements of the phased array ultrasonic probe are divided into five groups. The first group includes elements 1-50, which are used to scan the damage at the right upper side of the rail head. The second group includes elements 51-60, which are used to scan the damage near the right rail jaw of the steel rail. The third group includes elements 61-68, which are used to scan the damage at the middle part of the rail head, the rail waist, and the middle part of the rail bottom. The fourth group includes elements 69-78, which are used to scan the damage near the left rail jaw of the steel rail. The fifth group includes elements 79-128, which are used to scan the damage at the left upper side of the rail head. The wave form inside the rail is a transverse wave. The prior art detects the damage in the rail head, rail waist, rail jaw, and rail bottom area by simultaneously exciting multiple beams under the condition of water immersion. It can be seen that the prior art still has the following problems:

[0006] 1. The detection object and range of the prior art have deficiencies: the detection range of the prior art is only one rail top detection route, which belongs to the detection of internal defects of the steel rail. The 128-element probe in the prior art only detects the internal area of the rail head and part of the rail waist area, and cannot detect the entire rail bottom area and the weld excess height defect, which is not mentioned in the current standard;

[0007] 2. The prior art has limited detection and evaluation ability for small defects using phased array fan scanning technology: the prior art uses a 128-element phased array probe to perform fan scanning imaging based on the delay rule. The detection and evaluation ability for small defects is not strong, and the image expression ability for defects is not strong, which cannot directly and quickly locate the defects.

[0008] 3. The prior art has limited application range and low automation degree: the prior art is based on water coupling, which requires placing the steel rail inside a water tank, and the mechanical arm installation method is not proposed, only remaining in the defect detection of the steel rail before leaving the factory, which is not suitable for the detection needs of the actual service process of the steel rail, and the detection method has great limitations;

[0009] 4. The detection precision and accuracy of the prior art are highly interfered: in the actual detection process, the steel rail will have position deviation and deflection error relative to the mechanical arm, which seriously interferes with the precision and accuracy of the fixed scanning path method of the mechanical arm, and the prior art has not proposed a method to solve the position error, so the detection precision and accuracy are highly interfered.

[0010] The background description provided herein is for the purpose of generally presenting the context of the disclosure. The material described herein in this section is not prior art to the claims of the present application unless specifically indicated otherwise, and should not be admitted to be prior art by inclusion in this section. SUMMARY

[0011] In order to overcome the deficiencies in the background art, the present application discloses an ultrasonic phased array full-focus mechanical arm scanning method for detecting defects of a seamless steel rail weld.

[0012] In order to achieve the above-mentioned application purpose, the present application adopts the following technical solutions:

[0013] An ultrasonic phased array full-focus mechanical arm scanning method for detecting defects of a seamless steel rail weld, comprising the following steps:

[0014] S1, the flaw detection vehicle travels on the rail line, and the laser scanner synchronously scans the rail, and the flaw detection vehicle stops when the laser scanner collects the rail weld crown information; the flaw detection vehicle is loaded with a laser scanner and a mechanical arm;

[0015] S2, the host computer calculates the deflection and deviation of the current weld relative to the ideal weld position according to the three-dimensional coordinate data of the rail weld collected by the laser scanner through an automatic guidance positioning algorithm, and corrects the preset scanning path point set to obtain a new scanning path point set to generate a maintenance path;

[0016] The automatic guidance positioning algorithm is specifically: the three-dimensional coordinate data of the rail weld collected by the laser scanner is filtered and denoised through a point cloud voxelization algorithm, and then the deflection matrix of the relative deflection and the deviation matrix of the relative deviation are obtained through a point cloud matching algorithm according to the original point cloud after denoising, and the coordinates of each point in the preset scanning path point set are transformed according to the above-mentioned matrix to obtain the corrected new scanning path point set;

[0017] The preset scanning path includes: a first path for scanning a rail head, a rail waist and a rail bottom center region near a weld; a second path for scanning a rail waist region near the weld; a third path for scanning a rail waist and a rail bottom transition region near the weld; and a fourth path for scanning a weld crown side bottom surface region.

[0018] S3, the host computer drives the mechanical arm to clamp the ultrasonic phased array probe to scan along the inspection path;

[0019] S4, the ultrasonic phased array displays a real-time full-focus imaging image according to the signal data collected by scanning and can record the detected defects.

[0020] Specifically, the point cloud voxelization algorithm specifically includes the following steps:

[0021] T1, determine the maximum and minimum values of the point cloud data in the X, Y and Z axis directions;

[0022] T2, based on the maximum and minimum values in the X, Y and Z axis directions and the preset voxel grid size, establish a voxel grid and assign an index to it;

[0023] T3, for each voxel grid, determine whether the number of points in the grid is less than a preset screening threshold thV; when the number of points in the voxel grid is less than the screening threshold thV, discard all points in the voxel grid;

[0024] T4, for each remaining voxel grid, calculate the density value of the voxel grid according to the number of points in its surrounding neighborhood voxel grid and its own grid;

[0025] T5, output the voxel grid with the density value.

[0026] Specifically, the point cloud matching algorithm specifically includes the following steps:

[0027] D1, initialize the translation matrix T as the unit matrix;

[0028] D2, initialize the rotation matrix R as the unit matrix;

[0029] D3, perform the following iteration process until the convergence condition is met or the maximum number of iterations is reached:

[0030] a. select sample points from the source point cloud source_cloud;

[0031] b. find the nearest corresponding point for each sample point in the target point cloud target_cloud;

[0032] c. calculate the best rigid body transformation, including translation and rotation, to minimize the distance between the corresponding points of the source point cloud and the target point cloud;

[0033] d. Update the position of the source point cloud (source_cloud) using the calculated translation matrix T and rotation matrix R;

[0034] e. Calculate the amount of transformation change in this iteration. If the amount of transformation change is less than the preset convergence threshold, the registration process is considered to have converged, and the iteration is terminated.

[0035] f. Increase the iteration count iter. If iter reaches the preset maximum iteration count max_iterations, then terminate the iteration.

[0036] D4. Output the final translation matrix T and rotation matrix R;

[0037] D5. Calculate the corrected scan path point set. Based on the above output results, the corrected scan path point set can be calculated using the following formula:

[0038] P_correction = T*R*P_preset

[0039] Wherein, P_correction is the corrected set of scan path points, and P_preset is the preset set of scan path points.

[0040] Specifically, the first path is a straight line that coincides with the center line of the top of the rail, and the probe direction is parallel to the length direction of the rail during detection.

[0041] Specifically, the second path is a straight detection path along the web of the rail and perpendicular to the bottom surface of the rail.

[0042] Specifically, the third path is a straight line on the surface of the rail base, starting at the detection location of the defect in the transition area between the rail web and the rail base next to the weld, with the probe facing the rail length at a preset angle, and ending at the visible edge structure of the rail base.

[0043] Specifically, the fourth path is a straight line on the bottom surface of the rail, with the initial point and detection direction being the detection position of the defect where the center of the weld reinforcement edge is located and the probe orientation, and the termination point being the detection position of the visible weld reinforcement edge structure wave and the probe orientation. The probe orientation changes uniformly during the movement.

[0044] Specifically, the maintenance path also includes a second path, a third path, and a fourth path that are symmetrically copied to the other three quadrants excluding the quadrant itself and then connected to generate the path;

[0045] The quadrant division is as follows: based on the "I" shaped structure on the side of the rail, the rail can be divided into two sides along the axis according to the width of the rail. Then, based on the rails at both ends connected by the weld, the rails are divided into two ends along the extension direction of the rail, thus dividing the entire area to be inspected into four quadrants.

[0046] Specifically, step S1 specifically comprises:

[0047] S11, data acquisition: the laser scanner continuously scans the rail surface during the travel of the detection vehicle to obtain real-time three-dimensional point cloud data;

[0048] S12, feature extraction: real-time processing of the three-dimensional point cloud data to extract the contour features of the rail surface;

[0049] S13, feature recognition: analyzing the contour features to identify features representing the weld reinforcement; specifically comprising:

[0050] S131, detecting the mutation of the rail surface contour;

[0051] S132, identifying abnormal shapes inconsistent with the standard rail contour;

[0052] S133, analyzing the density change of the point cloud data;

[0053] S134, threshold judgment: comparing the identified features with the preset weld reinforcement feature threshold;

[0054] S14, stop signal triggering: if the identified features exceed the preset threshold, the system determines that the weld reinforcement has been detected, and triggers the stop signal;

[0055] S15, detection vehicle stopping: after the control system in the upper computer receives the stop signal, it immediately executes the braking program to make the detection vehicle stop safely in the shortest distance;

[0056] S16, position confirmation: after the detection vehicle stops, the system performs an accurate scan again to confirm the position of the weld.

[0057] Specifically, the detection vehicle is also loaded with a water spraying device to spray water onto the surface of the rail to be measured to form a water film.

[0058] The application provides a seamless rail weld defect ultrasonic phased array full-focus mechanical arm scanning method, which comprises the following steps: S1, a flaw detection vehicle travels on a rail line, a laser scanner synchronously scans the rail, and the flaw detection vehicle stops when the laser scanner collects rail weld reinforcement information; the flaw detection vehicle is loaded with the laser scanner and the mechanical arm; S2, an upper computer calculates the deflection and offset of the current weld relative to the ideal weld position according to the three-dimensional coordinate data of the rail weld side collected by the laser scanner, corrects the preset scanning path point set to obtain a new scanning path point set to generate a maintenance path; S3, the upper computer drives the mechanical arm to clamp the ultrasonic phased array probe to scan along the maintenance path; and S4, the ultrasonic phased array displays real-time full-focus imaging according to the signal data collected by scanning and can record the detected defects. The scanning method provided by the application realizes automatic scanning of the full-face position of the rail weld defect, realizes detection of the rail bottom rail jaw weld reinforcement area compared with the existing detection, and overcomes the defect that the existing technology cannot scan the full face.

[0059] In addition, the application can identify the weld and correct the mechanical arm scanning path through the automatic guided positioning technology, solve the positioning error influence of the mechanical arm scanning program in the automatic application, ensure that the flaw detection vehicle and the rail have relative center offset and deflection during operation, and guarantee the accuracy and reproducibility of the automatic detection.

[0060] In addition, the full-face detection full-process detection time of the application through the automatic guided positioning technology is 3 minutes, while the average time of manual detection is 30 minutes, which has obvious advantages in efficiency compared with the traditional manual detection, avoids the interference of the experience factors and personal state of manual detection on the detection result, and effectively improves the detection efficiency and detection accuracy.

[0061] In addition, the application provides a complete flaw detection vehicle detection system which can run autonomously on a rail line, and the application only needs to use the water spraying device of the system to spray water on the surface of the rail to be detected to form a water film, so that the detection conditions are provided, the problem that the detection conditions of the prior art need to be detected by immersion and the detection range is limited is solved, and the convenience of detection is improved.

[0062] In addition, the application uses the phased array full-focus technology to detect and quantify the vertical crack defects with a radius of 4.5 mm near the rail weld and the transverse hole with a diameter of 2 mm, all the defects are detected in repeated tests, the accuracy of the quantitative effect is 5.4% higher than the result of the fan scanning data, and the detection accuracy is effectively provided.

[0063] In addition, the relative error between the actual running defect center detection amplitude of the application and the defect center detection amplitude under the preset detection process is within 8%, and the experimental error is effectively controlled. BRIEF DESCRIPTION OF DRAWINGS

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 This is a schematic diagram of the ultrasonic phased array full-focusing robotic arm scanning method for seamless rail weld defects according to an embodiment of the present invention;

[0066] Figure 2 This is a schematic diagram of the overall hardware structure of an ultrasonic phased array fully focused robotic arm scanning method for seamless rail weld defects according to an embodiment of the present invention;

[0067] Figure 3 This is a schematic diagram of the main structure of the flaw detection vehicle according to an embodiment of the present invention;

[0068] Figure 4 This is a schematic diagram of the handle portion of the flaw detection vehicle provided according to an embodiment of the present invention;

[0069] Figure 5 This is a schematic diagram of the structure of the flaw detection vehicle mounting platform according to an embodiment of the present invention;

[0070] Figure 6 This is a schematic diagram of the structure of the flaw detection vehicle's height-adjustable frame according to an embodiment of the present invention;

[0071] Figure 7 This is a schematic diagram comparing noise reduction processing according to an embodiment of the present invention; wherein, Figure 7 (a) in the image shows the point cloud before denoising. Figure 7 (b) in the image shows the point cloud after denoising.

[0072] Figure 8 This is a schematic diagram of the detection path provided according to an embodiment of the present invention;

[0073] Figure 9 This is a schematic diagram of the first path detection area provided according to an embodiment of the present invention;

[0074] Figure 10 This is a schematic diagram of the first path provided according to an embodiment of the present invention;

[0075] Figure 11 This is a schematic diagram of the ultrasonic propagation path at the first path limit point provided by an embodiment of the present invention;

[0076] Figure 12is a second path detection area schematic diagram provided according to an embodiment of the present application;

[0077] Figure 13 is a second path schematic diagram provided according to an embodiment of the present application;

[0078] Figure 14 is a third path and fourth path detection area schematic diagram provided according to an embodiment of the present application;

[0079] Figure 15 is a third path schematic diagram provided according to an embodiment of the present application;

[0080] Figure 16 is a fourth path schematic diagram provided according to an embodiment of the present application;

[0081] Figure 17 is a whole scheme schematic diagram provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0082] The present application can be explained in detail through the following embodiments, the purpose of disclosing the present application is to protect all technical improvements within the scope of the present application, and in the description of the present application, it should be understood that if the orientation or position relationship indicated by the terms "upper", "lower", "front", "rear", "left", "right" and the like is only corresponding to the drawings of the present application, and is not intended to indicate or imply that the device or element must have a particular orientation.

[0083] Embodiment one

[0084] Reference Figure 1 The present embodiment provides an ultrasonic phased array full focusing mechanical arm scanning method for seamless rail weld defects, specifically comprising the following steps:

[0085] S1, the flaw detection vehicle travels on the rail line, and the laser scanner synchronously scans the rail, and the flaw detection vehicle stops when the laser scanner collects the rail weld reinforcement;

[0086] The mechanical arm end holds a 32-chip array element ultrasonic phased array probe; the ultrasonic phased array probe is loaded on a 36° wedge;

[0087] The present embodiment uses a self-designed flaw detection vehicle that can travel on the rail, and detects the defects near the rail weld, especially the crack defects, by means of the mechanical arm of the flaw detection vehicle holding a 32-chip array element ultrasonic phased array probe. The phased array probe is loaded on a 36° wedge to generate refracted transverse waves in the rail for defect detection, and 1024 groups of signals collected by 32 chips for excitation and reception are used for full focusing imaging to display the defect waveform.

[0088] Reference Figure 2 ,Figure 17 The overall structure of the mechanical arm and phased array is integrated on the designed inspection vehicle. The water tank for water supply and the generator for power supply are placed on the inspection vehicle. The generator is responsible for the power supply of the whole vehicle hardware. The overall control of the mechanical arm control cabinet, ultrasonic phased array and laser scanner is completed by the notebook computer. When the inspection vehicle travels on the railway track, the laser scanner scans the rail synchronously. When the laser scanner collects the key feature information of the rail weld excess height, the inspection vehicle stops immediately, and the laser scanner imports the three-dimensional coordinate data of the rail weld side into the host computer.

[0089] The overall hardware structure of the detection method in the embodiment is shown in Figure 2 , wherein the structure of the inspection vehicle includes the following blocks:

[0090] 1) Main block

[0091] Considering that the experimental equipment can be conveniently and flexibly used on the actual rail, the appearance structure design of the detachable hand-push inspection vehicle is carried out, the position layout of each instrument on the vehicle surface is planned, considering the weight of the vehicle itself, the simulation strength check of the vehicle is carried out using aluminum alloy and other materials, and aluminum alloy is selected as the vehicle body material considering the strength and light weight. In order to facilitate loading and transportation, the main body of the vehicle is divided into left and right two parts, and the middle part is connected by bolts, as shown in Figure 3 .

[0092] 2) Handle block

[0093] Considering the weight of the vehicle, the size of the pushing force, the convenience of transportation and disassembly, and combining with ergonomics, the size and inclination angle of the handle and the disassembly method are continuously adjusted, and finally a 45 steel material with a diameter of 50 mm is selected, the handle is assembled by three parts, and can be disassembled respectively, and is connected with the vehicle body by bolts as shown in Figure 4 .

[0094] 3) Installation table block

[0095] Considering that the installation table needs to load the mechanical hand, the bearing is large, and the deformation degree requirement is relatively strict, the installation table is designed separately, the material is selected and the strength is checked. After investigation, it is decided to select Q235 as the material, increase the rib plate on the basis of the simply supported beam form, and reduce the deformation degree of the installation table to 0.01 mm level. The installation table adjustment is shown in Figure 5 .

[0096] 4) Height increasing frame block

[0097] In order to facilitate the observation of the measurement results during the measurement process, the height increasing frame is added to lift the result display screen, which is convenient for the tester to observe the data and measurement condition while pushing the vehicle, as shown in Figure 6 .

[0098] 5) Guardrail plate and overall model:

[0099] Considering the movement of the car during the measurement process, add a guardrail to each instrument to limit the movement of the instrument, and adjust the layout of the instrument on the car to make its center of gravity central.

[0100] The flaw detection vehicle in this embodiment is manually driven, and can also be driven by loading an engine or an electric motor, as long as the driving function can be realized, and is not limited here.

[0101] In this embodiment, laser profile scanning is used as a visual sensor, and a depth camera, a binocular camera, a laser radar, etc. can also be used as a visual sensor of the device, and are not limited here.

[0102] The mechanical arm in this example is an ABB-1200 six-axis mechanical arm for detection, and other mechanical arms can also be used for detection, which is not limited here.

[0103] Step S1 specifically includes:

[0104] S11, data acquisition: the laser scanner continuously scans the surface of the rail during the travel of the flaw detection vehicle, and obtains real-time three-dimensional point cloud data;

[0105] S12, feature extraction: real-time processing of the three-dimensional point cloud data to extract the profile features of the rail surface;

[0106] S13, feature recognition: analyzing the profile features to identify features representing the weld crown;

[0107] Specifically includes:

[0108] S131, detecting the sudden change of the rail surface profile;

[0109] S132, identifying abnormal shapes that do not conform to the standard rail profile;

[0110] S133, analyzing the density change of the point cloud data;

[0111] S134, threshold judgment: comparing the identified features with the preset weld crown feature threshold;

[0112] The preset weld crown feature threshold is determined based on one or more combinations of height difference, width, or shape features;

[0113] S14, stop signal triggering: if the identified features exceed the preset threshold, the system determines that the weld crown has been detected, and triggers a stop signal;

[0114] S15, the flaw detection vehicle stops: after the control system in the upper computer receives the stop signal, it immediately executes the braking program to make the flaw detection vehicle stop safely in the shortest distance.

[0115] To further ensure the accuracy of the scanning position, a position confirmation operation can be performed after the flaw detection vehicle stops.

[0116] S16. Location Confirmation: After the flaw detection vehicle stops, the system performs another precise scan to confirm the weld location, preparing for subsequent inspections.

[0117] S2. The host computer uses an automatic guidance and positioning algorithm to calculate the current deflection and offset of the weld seam relative to the ideal weld seam position based on the three-dimensional coordinate data of the side of the rail weld seam collected by the laser scanner. It then corrects the preset scanning path point set to obtain a new scanning path point set and generates the maintenance path.

[0118] The automatic guidance and positioning algorithm is as follows: the three-dimensional coordinate data of the rail weld seam collected by the laser scanner is filtered and denoised by the point cloud voxel meshing algorithm, and then the deflection matrix of relative deflection and the offset matrix of relative offset are obtained by the point cloud matching algorithm based on the denoised original point cloud. The coordinates of each point in the preset scanning path point set are transformed according to the above matrices to obtain the corrected new scanning path point set, thus completing the dynamic adjustment of the preset scanning path.

[0119] The ideal weld location is the position of the weld when the flaw detection vehicle stops at the required distance according to the planned preset scanning path and the center plane of the flaw detection vehicle is parallel to the extension direction of the rail; at this time, the position of the bottom midpoint of the rail weld in the robot arm base coordinate system is x=230mm, y=0mm, z=-295mm, and the rail guide is parallel to the axis of the robot arm base coordinate system.

[0120] It is understandable that a path point set refers to a set of discrete coordinate points that constitute a path, and these discrete coordinate points can be connected continuously to form a path. This is existing technology and will not be elaborated here.

[0121] To enable the ultrasonic inspection robotic arm system to scan and locate seamless rail welds according to the planned program, it is necessary to provide offset correction for the robotic arm programming. This ensures that the flaw detection vehicle and the rail will experience relative center offset and deflection during operation, thus guaranteeing the accuracy and reproducibility of automated inspection.

[0122] This method utilizes the SmartRay 2D laser profilometer to implement an automatic guided positioning method based on a 2D profile laser scanner. First, the acquired 3D coordinate data of the rail weld is imported into MATLAB. Then, a point cloud voxel meshing algorithm is used to filter and denoise the acquired data, simplifying the large and complex 3D point cloud data.

[0123] Specifically, a three-dimensional voxel grid is created by using the input point cloud data, and then in each voxel, the center of gravity of all points in the voxel is used to approximate the other points in the voxel, so that a center of gravity point is finally used to represent all points in the voxel. The size of the voxel needs to be determined according to the actual data. After denoising filtering, some irrelevant point cloud data is deleted, and the remaining point cloud coordinates are extracted, and the denoising process is completed. The denoising effect comparison is as shown in Figure 7 Figure 7 (a) is the point cloud denoising process before Figure 7 (b) is the point cloud denoising process after.

[0124] The point cloud voxel grid algorithm specifically includes the following steps:

[0125] T1, determining the maximum and minimum values of the point cloud data in the X-axis, Y-axis and Z-axis directions;

[0126] T2, based on the maximum and minimum values in the X-axis, Y-axis and Z-axis directions and the preset voxel grid size, establishing a voxel grid and assigning an index to it;

[0127] T3, for each voxel grid, judging whether the number of points in the grid is less than a preset screening threshold thV; when the number of points in the voxel grid is less than the screening threshold thV, discarding all points in the voxel grid;

[0128] T4, for each remaining voxel grid, calculating the density value of the voxel grid according to the number of points in its surrounding neighborhood voxel grid and the grid itself;

[0129] T5, outputting the voxel grid with the density value.

[0130] The point cloud matching algorithm specifically includes the following steps:

[0131] D1, initializing the translation matrix T as the unit matrix;

[0132] D2, initializing the rotation matrix R as the unit matrix;

[0133] D3, performing the following iteration process until the convergence condition is met or the maximum number of iterations is reached:

[0134] a. selecting sample points from the source point cloud source_cloud;

[0135] b. finding the nearest corresponding point for each sample point in the target point cloud target_cloud;

[0136] c. calculating the best rigid transformation, including translation and rotation, to minimize the distance between the source point cloud and the target point cloud corresponding points;

[0137] ​d. Update the position of the source point cloud source_cloud with the translation matrix T and the rotation matrix R calculated;

[0138] e. Calculate the transformation change of this iteration, if the change is less than the preset convergence threshold tolerance, consider that the registration process has converged, terminate the iteration;

[0139] f. Increase the iteration number iter, if iter reaches the preset maximum iteration number max_iterations, terminate the iteration;

[0140] D4, output the final translation matrix T and rotation matrix R;

[0141] D5, calculate the corrected scanning path point set, the corrected scanning path point set can be calculated according to the output result using the following formula:

[0142] P_corrected=T*R*P_pre-set

[0143] Where P_corrected is the corrected scanning path point set, and P_pre-set is the preset scanning path point set.

[0144] The source point cloud source_cloud, the target point cloud target_cloud, the iteration number max_iterations, and the convergence threshold tolerance are obtained by a point cloud matching algorithm to obtain the translation matrix T and the rotation matrix R for correcting the path.

[0145] The target point cloud is obtained by modeling the whole rail and weld and exporting point cloud data according to the ideal position of the rail relative to the mechanical arm.

[0146] The ideal situation is when the flaw detection vehicle stops at the required distance according to the planned preset scanning path, and the center plane of the flaw detection vehicle is parallel to the extension direction of the rail.

[0147] The iteration number max_iterations can be selected between 50 and 100, in order to ensure better matching accuracy and iteration rate, through experiments, when the iteration number is 80, the accuracy is better, and increasing the iteration number will not significantly improve the accuracy.

[0148] The convergence threshold tolerance is 0.001-0.01, which can be comprehensively considered according to the actual calculation time and matching accuracy, and the actual selection is 0.007, which has better calculation effect.

[0149] The point cloud data, the preset voxel grid size, and the screening threshold thV are calculated to obtain the voxel grid with density value according to the voxel grid filtering method;

[0150] The preset voxel grid size is 0.7-0.9, which needs to be adjusted according to the point cloud density of the actual data set and the required down-sampling degree. The collected point cloud data of the rail weld is relatively dense, and 0.8 is selected for better effect.

[0151] The screening threshold thV ranges from 0.01 to 0.1, and 0.05 is selected for calculation according to the actual processing situation.

[0152] This calculation process is applied to each point on the preset path to obtain a complete modified scanning path. This modified path is the new path that the mechanical arm actually needs to follow, that is, the maintenance path, which takes into account the deviation of the actual position and direction of the rail. The host computer uses this maintenance path to guide the mechanical arm to perform rail weld flaw detection, thereby realizing automatic guided positioning under the guidance of laser positioning.

[0153] The preset scanning path includes: a first path for scanning the rail head, rail waist and rail bottom center area near the weld; a second path for scanning the rail waist area near the weld; a third path for scanning the transition area between the rail waist and the rail bottom near the weld; and a fourth path for scanning the side bottom area of the weld.

[0154] The host computer calculates the actual distance between the mechanical arm and the weld from the three-dimensional coordinate data of the rail weld side collected by the laser scanner, modifies the existing scanning path according to the offset deflection data, and then realizes the real-time position and attitude adjustment of the detection path, ensuring the accuracy of subsequent operations. Then, the mechanical arm control cabinet receives the host computer instructions, drives the mechanical arm to hold the ultrasonic phased array probe, executes the preset programming file along the modified path, at the same time, the water pump extracts clean water from the water tank and sprays through the nozzle at the end of the mechanical arm, forming a uniform water film on the surface of the rail weld to realize the water coupling condition required for ultrasonic detection, and completing the detection of the rail weld.

[0155] To cooperate with the ultrasonic detection mechanical arm system to realize the scanning and positioning of the seamless rail weld according to the planned specified program, the offset correction for the mechanical arm programming needs to be provided to ensure the relative center offset and deflection of the flaw detection vehicle and the rail during operation, and to ensure the accuracy and reproducibility of automatic detection.

[0156] The moving path of the mechanical arm for rail weld defect detection and the corresponding defect area containing cracks that can be detected near the rail weld are as follows, avoiding interference collision areas such as rail welds and rail fasteners during movement:

[0157] In this embodiment, path 1 is the first path, path 2 is the second path, path 3 is the third path, and path 4 is the fourth path, as shown in Figure 8 ;

[0158] The first path is a straight line coinciding with the rail head center line, and the probe direction is parallel to the rail length direction during detection.

[0159] (1) The rail head, rail waist, and rail bottom center area near the weld are detected, as shown in Figure 9

[0160] Path 1 is a straight line coinciding with the rail head center line, and the probe direction is parallel to the rail length direction during detection, as shown in Figure 10 The rail is cut in half along the rail head center line from the middle, and the propagation path at the extreme end point can be drawn, as shown in Figure 11

[0161] In one possible implementation, the distance between path 1 and the weld edge 177 mm is calculated and reasonably extended, and the length of path 1 is 354 mm, with the rail head center line as the center and 177 mm on each side, and the probe is turned 180° when it passes through the weld center line.

[0162] The second path is a straight line detection path along the rail waist part perpendicular to the rail bottom surface.

[0163] (2) The rail waist area near the weld is detected, as shown in Figure 12

[0164] Because the defect is detected parallel to the opposite side of the rail, according to the detection principle, the detection path 2 at the rail waist can be obtained by simply extending the detection point to a straight line perpendicular to the bottom surface and fitting the rail waist, as shown in Figure 13 It should be noted that the rail waist part is a curved surface, but the curve is gentle, the wedge can walk in a straight line, and the wedge gap with the rail is small, which can be filled with water coupling agent.

[0165] The length of path 2 needs to consider the extreme position at both ends, and there is a size limit of the clamping wedge. According to the size of the clamping tool designed in the present application, the detection path needs to ensure that the tool and the rail head lower jaw surface and the rail bottom surface maintain a 5 mm gap. Through simulation in solidworks software, path 4 is obtained, which is 46.5 mm away from the weld edge, with a length of about 69 mm, and the vertical distance from the highest point and the lowest point to the rail bottom surface is 114 mm and 45 mm, respectively.

[0166] The third path is a straight line extending along the incident point and the rail length at a first preset angle in the transition area between the rail waist and the rail bottom next to the weld.

[0167] Specifically, the third path is a straight line on the rail bottom surface, with the starting point being the detection position of the defect in the transition area between the rail waist and the rail bottom next to the weld, the probe direction being kept at a first preset angle along the rail length, and the ending point being the visible rail bottom edge structure wave point.​​​

[0168] In this embodiment, the first preset included angle is 27.3°.

[0169] The fourth path is a straight line detection path formed by selecting a plurality of points on the weld reinforcement side bottom surface intersection line and deriving the incident points by using a simulation propagation path, so that the incident points fall on a straight line.

[0170] Specifically, the fourth path is a straight line on the rail bottom surface, the initial point and the detection direction are the detection position of the defect where the weld reinforcement edge center is located and the probe orientation, and the terminal point is the detection position of the visible weld reinforcement edge structure wave and the probe orientation.

[0171] (3) Detect the transition area between the weld seam rail waist and the rail bottom and the weld seam reinforcement side bottom surface area, as shown in Figure 14 .

[0172] The weld seam rail waist and rail bottom transition area can be detected at the rail bottom position, and the weld seam reinforcement side bottom surface area can be detected. When detecting at the rail bottom, the incident point is as close as possible to the outer edge of the rail, and the water coupling agent flow is increased to ensure that the gap between the wedge and the rail bottom is filled. At this time, the path is mainly straight, and the fluctuation of the curve on the rail bottom multi-curved surface can be approximately abstracted as straight line movement, and a spring system will be assisted in the actual robot clamping detection process.

[0173] ① The weld seam rail waist and rail bottom transition position area detection path 3, as shown in Figure 15 .

[0174] According to the propagation path diagram Figure 8 , it can be known that because the crack at this position is mainly detected, the range to be detected is very small, and the change in the orientation angle caused by the movement of the probe in space can be ignored. The incident point at the weld seam rail waist and rail bottom transition position area moves appropriately along the detection interface, and a straight line detection path 3 is obtained.

[0175] The path 3 is 40mm long, and the included angle with the length direction of the rail is about 27.3°. The distal end point is about 127mm away from the weld seam edge, and about 23.3mm away from the rail edge along the rail width direction. The proximal end point is about 164mm away from the weld seam edge, and about 4mm away from the rail edge along the rail width direction.

[0176] The path 3 is for detecting the lower transition position of one side of the weld seam, and can be symmetrically extended to the other three positions.

[0177] ② The weld seam reinforcement side bottom surface area detection path 4, as shown in Figure 16 .

[0178] For the position of the side bottom surface of the weld reinforcement, the first wave is used to detect the position close to the weld on the rail bottom. It is difficult to calculate the position of the incident point, so multiple points on the junction line of the side bottom surface of the reinforcement are selected to derive the incident point through the simulation propagation path. Considering that the position of the ultrasonic incident point and the wedge direction are relatively free, the incident point is tried to be made to fall on a straight line, and a straight line detection path 4 is obtained.

[0179] The length of the path 4 is about 17.2 mm, and the included angle with the length direction of the rail is about 22.2°. The distance from the far end point to the weld edge is about 32.1 mm, and the distance from the rail edge in the width direction of the rail is about 19.9 mm. The limit detection position is on the junction of the side bottom surface of the reinforcement, which is about 5.34 mm away from the weld edge. The included angle between the deflection direction of the probe and the length direction of the rail is about 42.8°. The distance from the near end point to the weld edge is about 16.2 mm, and the distance from the rail edge in the width direction of the rail is about 13.4 mm. The limit detection position is on the junction of the side bottom surface of the reinforcement, which is about 20 mm away from the weld edge. The included angle between the deflection direction of the probe and the length direction of the rail is about 25°. In the moving process, the moving path is relatively short, and the probe can be considered to be uniformly deflected from 25° to 42.8° when moving from the near end to the far end.

[0180] The path 4 is for detecting the transition position of the side bottom surface of one side of the weld. It can be symmetrically extended to the other three positions. It can be found from the farthest end data and combined with the weld structure that due to the protruding part of the weld reinforcement on the rail bottom, the positions on both sides of the reinforcement are difficult to be detected, and there is a monitoring blind area.

[0181] In the embodiment, according to the I-shaped structure of the rail side, the rail can be divided into two sides along the axis in the width direction of the rail (i.e. perpendicular to the width direction of the rail). Then, according to the two end rails connected by the weld in the extension direction of the rail (i.e. in the length direction of the rail), the whole to-be-detected region can be divided into four quadrants according to the above division method.

[0182] Full-face detection requires a detection range including all four quadrants, and the above detection paths except the first path are only for detecting a single quadrant range of a rail side and a weld side, so the existing paths need to be symmetrically repeated and connected.

[0183] Therefore, the detection path further includes a second path, a third path, and a fourth path symmetrically replicated to the other three quadrants except the own quadrant and connected to generate a path.

[0184] Through such symmetric arrangement, comprehensive detection of the weld and the surrounding area can be ensured, and the possibility of defect detection is improved.

[0185] The specific symmetry conditions are as follows:

[0186] ① Symmetry of the second path (four positions, the probe direction is perpendicular to the path and faces the weld):

[0187] Original position: one side rail waist.

[0188] Symmetrical position 1: the other side rail waist.

[0189] Symmetrical position 2: the rail waist on the same side of the other end of the weld.

[0190] Symmetrical position 3: the rail waist on the other side of the other end of the weld.

[0191] The symmetry of the second path needs to be detected at four positions, in addition to the original path at one side rail waist, and these paths need to be repeated on the other side of the rail and the other end of the weld. In this way, it is ensured that the rail waist area is comprehensively detected on both sides and both ends of the weld.

[0192] ②Symmetry of the third path (four positions, probe direction towards the weld):

[0193] Original position: the rail waist and rail bottom transition area beside the one side weld.

[0194] Symmetrical position 1: the rail waist and rail bottom transition area beside the other side weld.

[0195] Symmetrical position 2: the rail waist and rail bottom transition area on the same side of the other end of the weld.

[0196] Symmetrical position 3: the rail waist and rail bottom transition area on the other side of the other end of the weld.

[0197] The third path is detected at four positions, which are the rail waist and rail bottom transition areas on both sides of the weld, covering both sides of the rail and both ends of the weld, thereby ensuring comprehensive detection of the rail waist and rail bottom transition areas.

[0198] ③Symmetry of the fourth path (four positions, probe direction towards the weld crown):

[0199] Original position: the weld crown side bottom surface area beside the one side weld.

[0200] Symmetrical position 1: the weld crown side bottom surface area beside the other side weld.

[0201] Symmetrical position 2: the weld crown side bottom surface area on the same side of the other end of the weld.

[0202] Symmetrical position 3: the weld crown side bottom surface area on the other side of the other end of the weld.

[0203] The fourth path is detected at four positions, covering all possible defect positions of the weld crown side bottom surface area, ensuring the comprehensiveness of the detection.

[0204] The third path and the fourth path both need to be symmetrical to the other three positions.

[0205] In summary, the four detection paths are combined to detect the location of the main difficult defects of the aluminum hot welding joint, and the path distribution is shown in Figure 8 The mechanical arm is connected according to the planned detection path to repeatedly detect the full section near the steel rail weld. During the detection process, the ultrasonic phased array is in the full focusing mode, the detection mode is TT, and the lowest value of the display distribution amplitude can be controlled according to the existing gate, so that the defect position containing cracks in the moving process can be directly observed. The center position of the defect is the depth and width distance of the steel rail weld near the current probe position, which guides the positioning and repair of the defect, and the size of the defect can be easily measured on the full focusing image by the-6dB method.

[0206] The embodiment provides a specific implementation:

[0207] The steel rail weld test block processed with five vertical crack defects and two horizontal through-hole defects at the positions of rail head, rail waist, rail bottom and weld seam excess height is simulated for detection, and the vertical crack defect with a radius of 4.5 mm and the horizontal through-hole with a diameter of 2 mm are detected at one time according to the detection method of the application, and the detection rate is 100% in 5 repeated experiments.

[0208] In the field of steel rail weld detection, the application of ultrasonic phased array full focusing technology has obvious advantages. This technology can realize comprehensive scanning of different position weld areas without frequent replacement of probes, effectively solving the problem that the traditional ultrasonic detection method is difficult to touch in certain positions. More importantly, with the high sensitivity and accuracy of the full focusing technology, it can accurately identify and locate the small defects in the weld that are difficult to detect. With the development of industrial technology, the mechanical arm has been widely used, which can make people free from heavy labor and replace people in dangerous environments, and has high positioning accuracy. The combination of mechanical arm and steel rail weld detection can effectively reduce the dependence on people in detection, and promote the development of intelligent and automatic detection of steel rail welds.

[0209] S3, the host computer drives the mechanical arm to clamp the ultrasonic phased array probe to scan along the maintenance path;

[0210] S4, the ultrasonic phased array displays real-time full focusing imaging according to the signal data collected during scanning and can record the detected defects.

[0211] During the scanning process according to the maintenance path, the ultrasonic phased array displays real-time full focusing imaging according to the signal data collected during scanning and can record the detected defects.

[0212] The whole detection process of the detection method in the embodiment is controlled by the self-developed software of the upper computer in the mobile detection process. The software mainly includes three interfaces: laser positioning instrument control interface, mechanical arm control interface and phased array control interface. By using the winform function of C#, the control software is made through the secondary development of the phased array equipment, laser scanner and mechanical arm, which is simple, intuitive and easy to use.

[0213] The laser scanning interface has the functions of connecting the laser scanner, setting the laser parameters, displaying the laser scanning image and data processing.

[0214] The mechanical arm control interface has the functions of connecting the mechanical arm, controlling the motor of the mechanical arm, displaying the real-time motion state of the mechanical arm and selecting the path.

[0215] The ultrasonic phased array control interface has the functions of connecting the phased array, setting the scanning parameters and displaying the scanning waveform.

[0216] In the laser scanning module, the ZLDS200Connect function is used to connect the laser scanner, the ZLDS200WriteParams function is used to modify the laser parameters, and finally the ZLDS200GetResult function is used to obtain the laser scanning data. Through data processing, the center position of the weld is obtained. When the laser is positioned at the appropriate position, the trolley stops and the mechanical arm starts to scan the rail weld.

[0217] In the mechanical arm control module, the ABB.Robotices.Controllers library is added to obtain the control authority of the mechanical arm. The scanner.Scan() function is used to scan the available mechanical arm, and the Controller.Logon function is used to connect the mechanical arm. The controller.Rapid.GetRapidData() function is used to read the current state of the mechanical arm, and the Mastership.(controller.Rapid) function is used to modify the path of the mechanical arm. When the trolley stops at the expected position, the mechanical arm performs scanning.

[0218] In the ultrasonic phased array control module, the parameter modification and display of the ultrasonic phased array are realized through the communication between the upper computer and the ultrasonic board card. The UtilsParameter.Instance().Connect(OnConnectCallback) function is used to connect the board card, the HardwareBase.SetGlobalParams function is used to set the parameters, and finally the ReceiveData(NetData netData) function is used to receive the waveform data.

[0219] In this embodiment, the time for a single weld inspection across the entire cross section is 1.5 minutes, and the total inspection time (i.e., automatic guidance and scanning) is 3 minutes, while the average time for manual inspection is 30 minutes. Compared with traditional manual inspection, this method has a significant advantage in efficiency and avoids the interference of experience factors and personal conditions on the inspection results.

[0220] This embodiment provides an ultrasonic phased array full-focus robotic arm scanning method for seamless rail weld defects, comprising the following steps: S1, a flaw detection vehicle travels along the rail track, and a laser scanner scans the rail synchronously. The flaw detection vehicle stops when the laser scanner acquires the rail weld excess height information; the flaw detection vehicle is equipped with a laser scanner and a robotic arm; S2, the host computer calculates the current deflection and offset of the weld relative to the ideal weld position based on the three-dimensional coordinate data of the rail weld side acquired by the laser scanner using an automatic guidance positioning algorithm, and corrects the preset scanning path point set to obtain a new scanning path point set to generate a maintenance path; S3, the host computer drives the robotic arm to hold the ultrasonic phased array probe and scan along the maintenance path; S4, the ultrasonic phased array displays a real-time full-focus imaging image based on the scanned signal data and can record the detected defects. The scanning method provided in this embodiment realizes automated scanning of the full cross-sectional location of rail weld defects. Compared with existing detection methods, it realizes the detection of the rail bottom and rail jaw weld excess height area, overcoming the shortcomings of existing technologies that cannot perform full cross-sectional scanning.

[0221] Furthermore, this embodiment uses automatic guidance and positioning technology to identify weld seams and correct the robotic arm's scanning path, thus solving the problem of positioning errors in robotic arm scanning programs during automated applications. This ensures that the flaw detection vehicle and the rail do not experience relative center offset or deflection during operation, guaranteeing the accuracy and reproducibility of automated detection.

[0222] This embodiment uses automatic guidance and positioning technology to experiment with different rail tilt and rotation conditions. It can successfully correct the detection path and continuously repeat the detection in the experiment. There were no cases where the test block deviated significantly from the preset detection position of the automatic detection process and thus failed to detect the problem.

[0223] Furthermore, the automatic guided positioning technology provided in this embodiment has a full-section inspection time of 3 minutes, while the average time for manual inspection is 30 minutes. Compared with traditional manual inspection, it has a significant advantage in efficiency and avoids the interference of experience factors and personal state on the inspection results, effectively improving inspection efficiency and accuracy.

[0224] In addition, the embodiment provides a complete flaw detection vehicle detection system, which can autonomously run on a track line, only needs to spray water to the surface of the steel rail to be detected by using a water spraying device of the system to form a water film, and can provide a detection condition, solves the problem that the detection condition of the prior art needs to be immersed in water for detection and the detection range is limited, and improves the convenience of detection.

[0225] In addition, the embodiment uses the phased array full focusing technology to detect and quantify the vertical crack defects with a radius of 4.5 mm near the steel rail weld and the transverse through hole with a diameter of 2 mm, the defects are all detected in repeated tests, the accuracy of the quantitative effect is 5.4% higher than the result of the fan-shaped scanning data, and the detection accuracy is effectively provided.

[0226] The relative error of the center detection amplitude of each defect in the actual operation of the embodiment and the center detection amplitude of the defect under the preset detection process is within 8%, and the experimental error is effectively controlled.

[0227] The specification is described with reference to flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one block or multiple blocks.

[0228] These computer program instructions can also be stored in a computer readable memory that can guide the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one block or multiple blocks.

[0229] These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide a process for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one flow or multiple flows and / or blocks Figure 1 The device that implements the functions specified in one block or multiple blocks.

[0230] The present application is not limited to the details of the foregoing exemplary embodiments and can be practiced with modification and alteration within the scope of the appended claims. Thus, the application is not to be limited to the details shown, as particularly provided in the above description, as these could vary in many ways.

Claims

1. An ultrasonic phased array fully focusing robotic arm scanning method for weld defects in seamless steel rails, characterized in that, Includes the following steps: S1. The flaw detection vehicle moves along the rail track, and the laser scanner scans the rail synchronously. The flaw detection vehicle stops when the laser scanner collects information on the weld height of the rail. The flaw detection vehicle is equipped with a laser scanner and a robotic arm. S2. The host computer uses an automatic guidance and positioning algorithm to calculate the current deflection and offset of the weld seam relative to the ideal weld seam position based on the three-dimensional coordinate data of the side of the rail weld seam collected by the laser scanner. It then corrects the preset scanning path point set to obtain a new scanning path point set and generates the maintenance path. The automatic guidance and positioning algorithm is as follows: the three-dimensional coordinate data of the rail weld seam collected by the laser scanner is filtered and denoised by the point cloud voxel meshing algorithm, and then the deflection matrix of relative deflection and the offset matrix of relative offset are obtained by the point cloud matching algorithm based on the denoised original point cloud. The coordinates of each point in the preset scanning path point set are transformed according to the above matrices to obtain the corrected new scanning path point set. The preset scanning paths include: a first path for scanning the central area of ​​the rail head, rail web, and rail bottom near the weld; a second path for scanning the rail web area near the weld; a third path for scanning the transition area between the rail web and rail bottom beside the weld; and a fourth path for scanning the bottom surface area of ​​the weld reinforcement side. The inspection path also includes a second path, a third path, and a fourth path symmetrically copied to the other three quadrants excluding the original quadrant and connected to generate a path; the quadrant division is as follows: based on the "I"-shaped structure on the side of the rail, the rail can be divided into two sides along the axis according to the width of the rail, and then the rails at both ends connected by the weld are divided into two ends along the extension direction of the rail, dividing the entire area to be inspected into four quadrants; the fourth path is a straight line on the bottom surface of the rail, with the initial point and the detection direction being the detection position of the defect where the center of the weld reinforcement edge is located and the probe orientation, and the ending point being the detection position of the visible weld reinforcement edge structure wave and the probe orientation, with the probe orientation changing uniformly during the movement. S3. The host computer drives the robotic arm to hold the ultrasonic phased array probe and scan along the maintenance path. S4. The ultrasonic phased array displays real-time full-focus imaging data based on the signals acquired during scanning and can record detected defects.

2. The robotic arm scanning method according to claim 1, characterized in that, The point cloud voxel meshing algorithm specifically includes the following steps: T1. Determine the maximum and minimum values ​​of the point cloud data in the X, Y, and Z axis directions; T2. Based on the maximum and minimum values ​​in the X, Y, and Z axes and the preset voxel grid size, create a voxel grid and assign indices to it; T3. For each voxel grid, determine whether the number of points in the grid is less than the preset filtering threshold thV; when the number of points in the voxel grid is less than the filtering threshold thV, discard all points in that voxel grid. T4. For each remaining voxel grid, calculate the density value of the voxel grid based on the number of points in its surrounding neighboring voxel grids and its own grid. T5 outputs a voxel mesh with density values.

3. The robotic arm scanning method according to claim 1, characterized in that, The point cloud matching algorithm specifically includes the following steps: D1. Initialize the translation matrix T as the identity matrix; D2. Initialize the rotation matrix R as the identity matrix; D3. Perform the following iterative process until the convergence condition is met or the maximum number of iterations is reached: a. Select sample points from the source point cloud; b. Find the nearest corresponding point for each sample point in the target point cloud target_cloud; c. Calculate the optimal rigid body transformation, including translation and rotation, to minimize the distance between corresponding points in the source point cloud and the target point cloud; d. Update the position of the source point cloud (source_cloud) using the calculated translation matrix T and rotation matrix R; e. Calculate the amount of transformation change in this iteration. If the amount of transformation change is less than the preset convergence threshold, the registration process is considered to have converged, and the iteration is terminated. f. Increase the iteration count iter. If iter reaches the preset maximum iteration count max_iterations, then terminate the iteration. D4. Output the final translation matrix T and rotation matrix R; D5. Calculate the corrected scan path point set. Based on the above output results, the corrected scan path point set can be calculated using the following formula: P_correction = T * R * P_preset Wherein, P_correction is the corrected set of scan path points, and P_preset is the preset set of scan path points.

4. The robotic arm scanning method according to claim 1, characterized in that, The first path is a straight line that coincides with the center line of the top of the rail, and the probe direction is parallel to the length direction of the rail during detection.

5. The robotic arm scanning method according to claim 1, characterized in that, The second path is a straight detection path along the web of the rail and perpendicular to the bottom surface of the rail.

6. The robotic arm scanning method according to claim 1, characterized in that, The third path is a straight line on the surface of the rail base, starting at the defect detection location in the transition area between the rail web and the rail base next to the weld, with the probe facing the rail length at a preset angle, and ending at the visible edge structure of the rail base.

7. The robotic arm scanning method according to claim 1, characterized in that, Step S1 specifically includes: S11. Data Acquisition: The laser scanner continuously scans the surface of the rail during the movement of the flaw detection vehicle to obtain real-time three-dimensional point cloud data; S12. Feature extraction: Process the three-dimensional point cloud data in real time and extract the contour features of the rail surface; S13. Feature Recognition: Analyze the contour features to identify features representing weld reinforcement height; specifically including: S131. Detect abrupt changes in the surface profile of the rail; S132. Identify abnormal shapes that do not conform to the standard rail profile; S133. Analyze the density changes of point cloud data; S134, Threshold Judgment: Compare the identified features with the preset weld reinforcement feature threshold; S14. Stop signal trigger: If the detected feature exceeds the preset threshold, the system determines that the weld excess height has been detected and triggers a stop signal. S15. Flaw detection vehicle stops: After receiving the stop signal, the control system in the host computer immediately executes the braking procedure to bring the flaw detection vehicle to a safe stop within the shortest distance. S16. Location Confirmation: After the flaw detection vehicle stops, the system performs another precise scan to confirm the weld location.

8. The robotic arm scanning method according to claim 1, characterized in that, The flaw detection vehicle is also equipped with a water spraying device to spray water onto the surface of the rail to be tested to form a water film.

Images