A rail weld grinding system

Abstract

The application discloses a steel rail weld polishing system, which comprises a polishing house and a control system, and is characterized in that the polishing house is internally provided with a polishing system and a visual positioning system, the polishing system comprises a mechanical arm and an end effector, the end effector comprises a sand belt mechanism and a support, the support comprises a cross arm and two connecting arms located on the two sides of the cross arm, one sand belt mechanism is installed on each connecting arm, and the two sand belt mechanisms are opposite to each other. The application can automatically realize qualitative defect detection of the surface quality of the steel rail and quantitative detection of the machining allowance, and can replace manual measurement such as manual visual observation and a micrometer on a current long rail production line, so that the detection efficiency is high, and the detection accuracy is high.

Description

Technical Field

[0001] This invention belongs to the field of rail grinding technology, and in particular relates to a rail weld grinding system. Background Technology

[0002] The standard length of steel rails produced by the steel mill is 25 meters or 100 meters. After being transported to the rail welding base, these standard-length rails are welded together to form 500-meter-long seamless rails. These are then transported to the railway line for laying. During on-site laying, specialized equipment is used to weld these 500-meter-long rails sequentially to create a continuous seamless rail line. There must be no mechanical gaps in any part of the rails to ensure smooth train operation. During welding at the rail welding base, a weld is formed at the joint of every two rails. This weld is thicker than the rail surface, forming a weld bead that protrudes from the rail surface. This weld bead must be ground away to create a smooth weld. Only when the weld and the rails on either side of it meet industry standards for a smooth transition can a qualified weld be formed, ensuring smooth and safe train operation.

[0003] Current methods for inspecting rail surface quality either involve no inspection at all, or rely on visual inspection and manual measurement of the rail surface depth at arbitrary points using a micrometer. There are no mechanized or automated inspection solutions. Manual inspection cannot detect pits or cracks on the rail base material surface, cannot inspect grinding marks and grinding quality at the weld joints after grinding, and cannot accurately measure the machining allowance at the rail weld joints. Summary of the Invention

[0004] Firstly, a grinding method for complex curved surface components is provided. The grinding object of this method is a component with a complex curved surface, the surface of which is a complex curved surface formed by connecting a plane, a vertical plane and a circular arc cylindrical surface. There are at least two circular arc surfaces, and the radii of the two circular arc surfaces are different. The method is divided into transverse grinding and longitudinal grinding. Transverse grinding refers to the height direction perpendicular to the plane, and longitudinal grinding refers to the length direction parallel to the axial direction of the circular arc cylindrical surface.

[0005] First, a grinding wheel with a tensioned abrasive belt is used to continuously grind laterally along a grinding trajectory from the plane to the vertical plane and then to the arc surface. The grinding wheel is a cylindrical roller with a flexible sleeve. The abrasive belt is tensioned on the flexible sleeve. The grinding wheel is tangent to the structural component. While the grinding wheel rotates, it drives the abrasive belt to move, thereby achieving the removal of material from the surface of the structural component. During the lateral grinding process, the grinding wheel rotates and feeds along the grinding trajectory, and also has a longitudinal feed amount. After the lateral grinding is completed, longitudinal finishing grinding is performed. After longitudinal finishing grinding, the machining allowance formed between the grinding area and the non-grinding area is at least 0.5mm.

[0006] A circular arc cylinder refers to a cylindrical surface formed by stretching a circular arc along its length.

[0007] Furthermore, the structural member includes a top surface and a bottom surface, with at least one waist surface between the top and bottom surfaces. Between the top surface and the waist surface, there is a first vertical surface and a first transition arc surface. The first transition arc surface is below the first vertical surface, and in the top view direction, the first vertical surface serves as the boundary of the top surface. The first transition arc surface is within the coverage area of ​​the first vertical surface and the top surface. Between the waist surface and the bottom surface, there is a second vertical surface and a second transition arc surface. The second vertical surface is below the second transition arc surface, and in the bottom view direction, the second vertical surface serves as the boundary of the bottom surface. The second transition arc surface is within the coverage area of ​​the second vertical surface and the top surface. The maximum width of the structural member is 150 mm, and the maximum height is 176 mm.

[0008] As an application scenario, this method uses steel rails, such as those used in railways, subways, and light rails, as components with complex curved surfaces. The method involves online grinding of welded steel rails on a long rail production line using abrasive belts. The method is divided into transverse grinding and longitudinal grinding. Transverse grinding refers to grinding along the height of the rail, while longitudinal grinding refers to grinding along its length. During transverse grinding, the abrasive belt grinds along the first and second grinding tracks from both sides of the rail's width direction, using flexible force to polish and remove weld beads. The first and second grinding tracks correspond to one side of the rail web, with the first grinding track covering the top surface and the second grinding track covering the bottom surface, or vice versa. During longitudinal grinding, a grinding wheel is fed longitudinally along the rail.

[0009] The weld bead is divided into two grinding tracks, each grinding from top to bottom, achieving continuous automated grinding of the rail top, rail web, and rail bottom. The grinding unit uses flexible force to abut against the rail, avoiding the problem of hard damage caused by rigid force grinding. Furthermore, if the weld bead is too strong, the grinding unit will slip, preventing damage to the grinding unit itself.

[0010] Furthermore, during transverse grinding, the abrasive belt moves not only laterally along the rail but also longitudinally along the rail.

[0011] Furthermore, the sanding belt is part of the grinding unit, which includes a grinding wheel and a drive wheel. The sanding belt is tensioned between the grinding wheel and the drive wheel. The sanding belt, grinding wheel, and drive wheel form a belt drive mechanism. During transverse grinding, the axial direction of the grinding wheel is aligned with the length direction of the rail. The grinding wheel abuts against the rail, and the sanding belt is located between the grinding wheel and the rail. The grinding wheel rotates, and the sanding belt removes weld beads along the tangent of the grinding wheel.

[0012] Furthermore, during transverse grinding, the force between the grinding wheel and the rail is monitored. If the monitored force exceeds the set force threshold, an alarm is triggered.

[0013] Furthermore, during longitudinal grinding, the grinding includes grinding the rail web, grinding the rail top surface, and grinding the rail bottom surface. A flap wheel is used to grind the rail web, a first disc grinding wheel is used to grind the rail top surface, and the grinding surface of the disc angle grinding wheel is the bottom surface of the grinding wheel; a second disc grinding wheel is used to grind the rail bottom surface, and the grinding surface of the disc angle grinding wheel is the bottom surface of the grinding wheel.

[0014] Furthermore, a first disc grinding wheel is used to grind the curved surface transitioning from the rail web to the rail bottom, and a second disc grinding wheel is used to grind the curved surface below the rail top and transitioning to the rail web; the first disc grinding wheel forms a first angle with the horizontal plane, the angle of the first angle being 0° to -10°; the second disc grinding wheel forms a second angle with the horizontal plane, the angle of the second angle being 0° to 10°.

[0015] Furthermore, the second disc grinding wheel has a first grinding section and a second grinding section, which are concentric, with the first grinding section inside and the second grinding section outside. Viewed from the axial section of the grinding wheel, the second grinding section is arc-shaped, and its central angle on the axial section coincides with the central angle of the curved surface below the rail top and transitioning to the rail web. This coincidence does not refer to mathematical equality, but rather that the central angle of the second grinding section is close to, but less than or equal to, the central angle of the curved surface.

[0016] Furthermore, the part of the grinding wheel that contacts the abrasive belt is a flexible component. When grinding the top surface of the rail, the grinding wheel rotates and feeds towards the centerline in the direction of rail width while simultaneously moving from one side of the rail web to the other. At the same time, the grinding wheel reciprocates longitudinally along the rail, with a lateral feed distance of not less than 35mm. When grinding the rail web, the grinding wheel rotates and feeds towards the centerline in the direction of rail width while simultaneously feeding along the curved surface of the rail web in the height direction. At the same time, the grinding wheel reciprocates longitudinally along the rail, with a lateral feed distance of not less than 35mm. This limitation on the lateral feed distance ensures that the abrasive belt only grinds away weld beads without damaging the rail base material.

[0017] Furthermore, the lateral feed distance is 3.5 cm to 5 cm. At this lateral feed distance, weld beads are removed cleanly without damaging the rail base material.

[0018] Furthermore, after the welded rails undergo misalignment testing, the weld beads of rails with acceptable misalignment are ground down using this grinding method.

[0019] Furthermore, before grinding, data on the misalignment of the rail is obtained to identify the higher base material on both sides of the rail weld. The grinding wheel is then fed laterally from the higher base material to the lower base material. This achieves the goal of lateral grinding without damaging the rail base material.

[0020] Furthermore, before grinding, the height difference between the weld and the base material is measured, and the grinding wheel feeds towards the centerline in the direction of rail width by a margin less than this height difference. Optionally, this height difference can be the average height difference between the weld and the base material, or the minimum height difference between the weld and the base material, or the median height difference between the weld and the base material. This avoids damage to the rail base material caused by overfeeding.

[0021] Optionally, the belt drive mechanism, consisting of the sanding belt, grinding wheel, and drive wheel, is mounted on a bracket. During the grinding operation, if the sanding belt remains in contact with the rail for more than a set time threshold, the entire belt drive mechanism stops feeding towards the rail, and the bracket drives the grinding wheel to retract away from the rail until the force between the grinding wheel and the rail is reduced to within an acceptable range. This prevents damage or failure of the belt drive mechanism.

[0022] One scheme for retracting the grinding wheel away from the rail using a support structure is as follows: the support section for mounting the grinding wheel and the support section for mounting the drive wheel are independent of each other, and the two support sections are connected together by a flexible or elastic component. When the contact time between the sanding belt and the rail does not exceed a set time threshold, the flexible or elastic component keeps the relative position of the grinding wheel and the drive wheel stable, allowing the grinding wheel to perform the grinding operation. When the contact time between the sanding belt and the rail exceeds the set time threshold, the flexible or elastic component deforms, causing the grinding wheel to move closer to the drive wheel.

[0023] Another scheme, where the support moves the grinding wheel backward away from the rail, involves mounting the grinding wheel on a frame. An elastic or flexible element is installed between the support and the frame to create a flexible force during grinding. When the contact time between the sanding belt and the rail does not exceed a set time threshold, the flexible or elastic element stabilizes the relative position of the support and the frame, allowing the grinding wheel to perform the grinding operation. When the contact time exceeds the set time threshold, the flexible or elastic element deforms, causing the support to move closer to the frame. The frame can be connected to a mechanized or automated feed power unit to feed the grinding wheel.

[0024] These two types of brackets, which drive the grinding wheel to retract away from the rail, can be used independently or combined to form a composite flexible force mechanism.

[0025] Secondly, a rail grinding system includes a grinding unit, a rail support mechanism, and a feeding mechanism for feeding the grinding unit. The feeding mechanism is connected to the grinding unit, and a buffer assembly is provided between the grinding unit and the feeding mechanism. The buffer assembly includes a spring or a flexible component other than a spring. The grinding unit includes a grinding wheel, a drive wheel, and a sanding belt. The sanding belt is tensioned between the grinding wheel and the drive wheel. The sanding belt, the grinding wheel, and the drive wheel form a belt drive mechanism. During transverse grinding, the axial direction of the grinding wheel is aligned with the length direction of the rail.

[0026] Furthermore, the grinding unit includes a mounting frame, with a grinding wheel and a drive wheel mounted on the mounting frame. A tension wheel is provided on the mounting frame and connected to a tensioning drive component. The tension wheel tensions the sanding belt on the grinding wheel and the drive wheel. During the grinding task, the grinding wheel rests against the rail. The grinding unit has a damping mechanism that limits the force threshold of the grinding wheel. When the force between the grinding wheel and the rail exceeds the force threshold, the grinding wheel displaces towards the drive wheel.

[0027] Furthermore, the grinding wheel includes a rotating shaft and a flexible sleeve, with the rotating shaft inside and the flexible sleeve outside, and the sanding belt in contact with the flexible sleeve.

[0028] Furthermore, the mounting frame is provided with a tensioning drive and a tensioning wheel. The tensioning drive includes a fixed part and a movable part. The tensioning wheel is located in the movable part, and the fixed part is located in the mounting frame.

[0029] Furthermore, the mounting frame is equipped with a limiting roller, the axis of which is parallel to the axis of the grinding wheel; when the limiting roller touches the rail, the grinding wheel reaches the limit position of its feed into the rail.

[0030] Furthermore, the mounting frame includes a fixed part and a movable part, with the grinding wheel located in the movable part and the transmission wheel located in the fixed part; when the force between the grinding wheel and the rail is greater than the force threshold, the movable part retracts; when the force between the grinding wheel and the rail is less than the force threshold, the relative positions of the movable part and the fixed part are fixed.

[0031] Furthermore, the mounting bracket is equipped with a damping mechanism, which includes a spring and a locking element. The spring is positioned between the movable part and the fixed part. The spring and the locking element work together to keep the movable part in a balanced position, and the distance between the grinding wheel and the transmission wheel is fixed.

[0032] Furthermore, the fixed part is provided with a first spring mounting component, and the movable part is provided with a second spring mounting component, with the two ends of the spring being fixed to the first spring mounting component and the second spring mounting component, respectively.

[0033] Optionally, the centers of the grinding wheel and the drive wheel are positioned differently, one higher and one lower, with the tensioning wheel located between the grinding wheel and the drive wheel. The triangle formed by the centers of the grinding wheel, the drive wheel, and the tensioning wheel lies within the area enclosed by the sanding belt. This configuration provides good tension stability of the sanding belt and allows for adjustable friction between the sanding belt and the grinding wheel, which helps maintain a stable relative position between them even when the grinding wheel is rotating at high speed.

[0034] Furthermore, the tensioning drive component is a cylinder, hydraulic cylinder, or electric push rod. The tensioning drive component intersects the straight line formed by the center of the grinding wheel and the center of the transmission wheel. The tensioning wheel is closer to the transmission wheel than the grinding wheel.

[0035] Furthermore, when the sanding belt is tensioned, the section of the sanding belt between the grinding wheel and the drive wheel forms an angle with the horizontal plane, the section of the sanding belt between the grinding wheel and the tensioning wheel forms an acute angle, and the section of the sanding belt between the drive wheel and the tensioning wheel forms an obtuse angle with the section of the sanding belt between the grinding wheel and the tensioning wheel. This maintains the stability of the grinding unit and the sanding belt.

[0036] Furthermore, acute angles are less than 45°, and obtuse angles are less than 150° but greater than 120°.

[0037] Furthermore, a first reference plane is constructed, passing through the axis of the grinding wheel and the axis of the transmission wheel. The center plane of the grinding wheel in the width direction is used as the second reference plane, which is perpendicular to the first reference plane. The intersection of the second reference plane and the axis of the tensioning wheel is projected onto the line of intersection of the first and second reference planes to obtain the tensioning wheel projection point. The point of intersection of the axis of the grinding wheel and the first and second reference planes is used as the grinding wheel projection point, and the point of intersection of the axis of the transmission wheel and the first and second reference planes is used as the transmission wheel projection point. The ratio of the distance between the tensioning wheel projection point and the grinding wheel projection point to the distance between the tensioning wheel projection point and the transmission wheel projection point is at least 5:4.

[0038] Optionally, the rail grinding system has two grinding units as described above, with the grinding units respectively located on both sides of the rail width direction, one grinding unit being higher and the other grinding unit being lower.

[0039] Furthermore, the tension wheel of one grinding unit faces upwards, while the tension wheel of the other grinding unit faces downwards.

[0040] Furthermore, the polishing system includes a bracket for mounting polishing units, with two polishing units mounted on the same bracket; the bracket includes a horizontal arm and two connecting arms located on both sides of the horizontal arm, with one polishing unit mounted on each connecting arm, and an elastic mechanism between the connecting arm and the horizontal arm; the horizontal arm is provided with an electrical connector for connecting to the robot.

[0041] Optionally, the ratio of the distance from the center of the grinding wheel of the high-grinding unit to the distance from the center of the grinding wheel of the low-grinding unit to the horizontal arm is 0.7 to 0.9. With this setting, when switching from grinding one side of the rail to the other side, the pose of the robotic arm or robot itself remains basically unchanged; only the pose of the support needs to be switched.

[0042] Furthermore, the connecting arm includes a slant plate connected to the cross arm and a tripod connected to the slant plate. The mounting bracket of the grinding unit is connected to the slant plate and the tripod, and is located at the bottom of the slant plate and the tripod. The mounting bracket, slant plate, and tripod form a triangular structure. The center of the drive wheel is located within the area covered by the triangular structure, and the drive wheel is located outside the triangular structure. A flexible mechanism can be provided between the slant plate and the cross arm, such as a damper, allowing the slant plate and the cross arm to have millimeter-level rotational displacement. The damper can be a spring, a rubber pad, a toothed connector, etc. A flexible mechanism can also be provided between the tripod and the slant bar. The rigidity (or flexibility) of the connection between the tripod and the slant plate can be different from the rigidity (or flexibility) of the connection between the slant plate and the cross arm, giving the entire support multiple flexible mechanisms.

[0043] Furthermore, the movable part of the mounting bracket is located outside the triangular structure, and the fixed part is part of the triangular structure; the mounting bracket is provided with a limit block, and the triangular bracket achieves blocking and limiting of the limit block.

[0044] Furthermore, the connecting arm is equipped with a cooling medium pipe and a cooling control valve. The cooling medium pipe extends from the inclined plate towards the tripod, and the tripod is equipped with a positioning element for the cooling medium pipe. The cooling medium pipe includes a rigid pipe and a flexible pipe. The rigid pipe is connected to the positioning element, and the flexible pipe is connected to the cooling control valve and the rigid pipe.

[0045] Optionally, the feeding mechanism is a robotic arm with a quick-change interface on the horizontal arm. The connection surface between the quick-change interface and the robotic arm is inclined relative to the horizontal arm.

[0046] Optionally, the feeding mechanism is a three-dimensional moving platform, with the center of the cross arm connected to the three-dimensional moving platform. The three-degree-of-freedom moving platform provides translational displacement of the support in the X, Y, and Z axes.

[0047] Optionally, the feeding mechanism is a three-dimensional moving platform, which is connected to the mounting frame of each grinding unit. In this case, the grinding unit is directly connected to the three-dimensional moving platform, eliminating the need for a support structure. For example, the three-dimensional moving platform uses a plane parallel to the bottom of the rail as the XOY plane, the length direction of the rail as the X-axis, and the height direction of the rail as the Z-axis. The three-dimensional moving platform enables the sanding belt mechanism to translate along the x, y, and z axes to achieve grinding feed; each grinding unit is mounted on its own Z-axis translation module, and the two grinding units share the x and y-axis translation modules.

[0048] Optionally, the aforementioned rail grinding system is installed inside a grinding chamber, which isolates the grinding work area from the external environment, enabling unmanned grinding operations.

[0049] Optionally, the grinding chamber has a rotating wall that can rotate around its centerline. There is a sealing mechanism between the boundary of the rotating wall and the grinding chamber. The rotating wall is set on a turntable, the center of which is connected to a turntable motor. A pair of shelves are provided on the turntable, with the two shelves located on opposite sides of the rotating wall.

[0050] In this way, it is possible to replace worn sanding belts, grinding wheels and other consumable parts or carry out maintenance outside the grinding room without stopping the machine inside.

[0051] Optionally, the tensioning drive is a tensioning cylinder. The shelf is equipped with an air supply port, and the sanding belt mechanism is equipped with an air inlet. When the support of the grinding unit is placed on the shelf, the air supply port is connected to the air inlet. The air supply port is connected to the air source. The shelf is equipped with a force sensor. When the force is sensed, the air source is turned on. When the force is zero, the air source is turned off.

[0052] The aforementioned rail grinding system is applied to long rail production lines, using polishing and grinding to remove weld beads from rail welds, forming a rail rough grinding workstation for long rail production lines.

[0053] A rail rough grinding workstation for a long rail production line, wherein the rail grinding system is installed on the long rail production line after the misalignment detection station after welding.

[0054] Thirdly, this invention provides an automated rail inspection system for detecting the surface quality of rails. For example, it performs quantitative and qualitative testing on the surface quality of polished rails.

[0055] An automated inspection device includes a camera, point light sources, and an inspection unit support. The camera and point light sources are mounted on the inspection unit support. A single camera and multiple point light sources configured for that camera constitute one inspection unit. The coverage areas of the light spots from the point light sources within the same inspection unit on the object under test are independent. During a single shooting task, the point light sources are sequentially illuminated, and the camera captures an image of the object under test each time a point light source illuminates. Only one point light source is illuminated during each shot. The positions of the point light sources and the camera remain relatively fixed during a single shooting task. Each shooting task acquires multiple images of the object under test. Each image corresponds to an image where a point light source illuminates the object, simultaneously containing object information and illumination information. The spatial features and inter-frame features of the object images are fused to obtain a three-dimensional image of the surface texture of the object under test.

[0056] Optionally, spatial features and inter-frame features are extracted from the image of the object to be tested. Both spatial features and inter-frame features are represented by three-dimensional convolutions. The three-dimensional convolution includes a two-dimensional spatial dimension and a one-dimensional inter-frame dimension. The value of the one-dimensional inter-frame dimension in the three-dimensional convolution of spatial features is a set value, and the two-dimensional spatial dimension is the spatial feature value of the image. The value of the two-dimensional spatial dimension in the three-dimensional convolution of inter-frame features is a set value, and the one-dimensional inter-frame dimension is the inter-frame feature value. When performing image information fusion, the inter-frame features are processed first, then the spatial features are processed, and then the inter-frame features and spatial features are fused to obtain a three-dimensional image of the surface texture of the object to be tested.

[0057] Optionally, the point light sources are controlled by a controller. The controller has preset lighting rules for the point light sources. Each point light source has its own unique light source code, and each camera has its own unique camera code. The lighting rules include the correspondence between camera codes and light source codes, as well as the lighting order of the point light sources. For example, camera 1 corresponds to light source 01, light source 04, and light source 06. Light sources 01, 04, and 06 are lit individually in sequence.

[0058] Alternatively, the point light source is fixed to the detection unit bracket via a light source bracket. Each light source bracket has one point light source. The light source bracket includes a hanging ear and a hanging piece. The hanging ear is fixed to the detection unit bracket, and the light source is fixed on the hanging piece. The hanging piece and the hanging ear are rotatably connected, and there is damping between the hanging piece and the hanging ear.

[0059] Optionally, the hook has an ear extending away from the detection unit support, and the hanger has a beam and a connecting part extending towards the detection unit support. The point light source is fixed to the beam; the connecting part overlaps with the ear, and there is a light source pivot between the connecting part and the ear. The light source pivot is fixed to the connecting part or the ear. Rotating the beam allows adjustment of the illumination area of ​​the light source on the object being measured.

[0060] Optionally, the light source shaft is fixed to the connecting part, and the light source shaft is connected to the output shaft of the light source drive motor. The presence of the light source shaft allows the position of the light source to be adjusted, thus adapting to objects of different sizes and specifications. Using an encoder or angle measuring instrument, the relationship between the size of the object being measured and the rotation angle of the light source can be obtained, facilitating rapid calibration of the light source position when changing the object being measured.

[0061] One possible approach is to group the light sources based on their distance from the camera, using the camera as the center and any plane perpendicular to the camera axis as the reference plane. Light sources in the same group should have the same distance from the camera, while light sources in different groups should have random distances from the camera. Each group should contain at least three light sources.

[0062] Optionally, the polar coordinate angle of the light source can be randomized with the camera as the center. In this way, the light source can illuminate the object being measured as discretely as possible.

[0063] Optionally, the ratio of the distance between any two adjacent light sources to the shooting distance from the camera to the object under test is 0.1 to 0.2. With this setting, the image captured by the camera is clear, and the illumination areas of the light sources on the object under test do not overlap, resulting in good 3D image reconstruction of the surface texture of the object under test.

[0064] During the production of long rails, the surface quality requirements for rails include: the weld seam and the weld slag around the weld seam must be cleaned, the ground surface must be smooth, the rail base (the curved part where the rail web transitions to the rail base) must be rounded, and lateral grinding is not allowed, and the rail base material must not be damaged.

[0065] When the aforementioned automated inspection device is applied to the inspection of rail surface quality, an automated rail inspection system is formed. An automated rail inspection system includes at least one set of the aforementioned inspection devices. The inspection object of this system is a rail located on a production line that has completed initial milling or initial grinding of weld beads. The inspection system is set after the first removal process of the rail weld, or there are two sets of the inspection system: the first set is set after rail misalignment detection and before the first removal process of the rail weld; the second set is set after the first removal process of the rail weld.

[0066] In some embodiments, the rail inspection system has a frame in which an automated inspection device is installed. The frame has side plates and a top plate that enclose the interior of the frame into a dark chamber. One side plate has an entrance that allows rails to enter, and the opposite side plate has an exit that allows rails to exit.

[0067] Optionally, the side panel may have a door or bed, or the side panel may be detachably connected to the frame.

[0068] In the production of long rails, the requirements for surface inspection of rails also include: full-section inspection of rail welds.

[0069] An alternative is to install a detection drive device within the frame that can move 360° around the rail cross-section. The detection device is connected to the detection drive device via an interface, and the detection drive device moves the detection device step by step to perform detection operations on each surface of the rail.

[0070] Optionally, there is only one detection device. The detection drive device moves in a step-by-step manner to perform detection operations on the top surface of the rail, the web surface of the rail, the bottom surface of the rail, and another web surface of the rail.

[0071] Alternatively, the frame may have an upper detection assembly and a lower detection assembly. The upper detection assembly may have a first detection device parallel to the ground, a second detection device located on both sides of the first detection device, and a third detection device. The second and third detection devices may be symmetrical about the camera center of the first detection device.

[0072] Optionally, the upper detection assembly is connected to the detection drive device. The second detection device detects one rail web surface, and the third detection device detects another rail web surface. The second and third detection devices each have three detection positions. The detection drive device causes the second or third detection device to move between the three detection positions. At the first detection position, the angle between the axis of the camera of the second detection device and the horizontal plane is -25° to -38°. At the second detection position, the axis of the second detection device is parallel to the horizontal plane. At the third detection position, the angle between the axis of the second detection device and the horizontal plane is 35° to 55°.

[0073] Optionally, the third detection device may not operate when the second detection device is in operation, or vice versa.

[0074] Optionally, the lower inspection component is fixed below the rail base.

[0075] An alternative approach is to have four sets of testing devices, each corresponding to a testing area for one rail. The testing devices form a testing area, and the rail enters the testing area along the axial direction of the prism. The distance between two opposing testing devices is no less than 1050mm. This allows the testing device to be suitable for rail testing on long rail production lines.

[0076] Optionally, one set of detection devices is aligned with the top surface of the rail, one set of detection devices is aligned with the bottom surface of the rail, one set of detection devices is aligned with one rail web surface, and another set of detection devices is aligned with another rail web surface. Each set of detection devices on the rail web surface has an upper detection unit, a middle detection unit, and a lower detection unit. Each detection unit has a camera, a point light source, and a light source controller. The camera axis of the upper detection unit is tilted downwards at an angle of 27° to 34° with the horizontal plane, the camera axis of the middle detection unit is parallel to the horizontal plane, and the camera axis of the lower detection unit is tilted upwards at an angle of 41° to 49° with the horizontal plane.

[0077] Optionally, any set of detection devices on the rail web surface includes a back plate. The upper detection unit, middle detection unit, and lower detection unit are respectively installed on the back plate through their respective position adjustment mechanisms. The back plate includes a main plate, an upper wing plate, and a lower wing plate. The main plate is parallel to the height direction of the rail to be tested. The upper wing plate intersects with the main plate, and the top of the upper wing plate is closer to the rail to be tested than the bottom of the upper wing plate. The lower wing plate intersects with the main plate, and the top of the lower wing plate is farther away from the rail to be tested than the bottom of the lower wing plate.

[0078] Optionally, the camera of the upper detection unit is located near the area where the upper wing plate and the main board intersect, a portion of the light source of the upper detection unit is located in the area covered by the upper wing plate, and another portion of the light source of the upper detection unit is located in the area covered by the main board.

[0079] Optionally, the camera and its light source of the detection unit are located within the area covered by the motherboard.

[0080] Optionally, the camera and its light source of the lower detection unit are located within the area covered by the lower wing plate.

[0081] In this way, any set of detection devices on the rail web surface can perform full surface detection of the surface of the rail web surface located below the rail top, the side surface of the rail web, the small arc surface transitioning between the surface below the rail top and the side surface of the rail web, the large arc surface transitioning from the rail web to the rail bottom, and the vertical surface adjacent to the rail web and the rail bottom.

[0082] In the production of long rails, there are also requirements for the inspection of the rail surface: the height of the rail bottom weld bead (weld) after grinding should not be less than 0.5mm. Therefore, it is also necessary to perform quantitative measurement on the ground rail.

[0083] An optional solution for the quantitative measurement device is: the detection device has a line laser, the axis of the line laser is at an angle to the axis of the camera, the line laser does not emit laser when the light source is lit, and the light source is not lit when the line laser emits laser lines.

[0084] Based on the requirements for rail surface quality, we specify the following indicators for detecting defects on the surface of the polished rail: whether there is a transverse polishing track, whether there is a step between the polished surface and the rail base material, and propose a method for detecting defects on the surface of rail using the above-mentioned detection device or system.

[0085] A method for detecting surface defects on rails using a camera and a group of point light sources is disclosed. This method uses a camera and multiple point light sources configured for that camera as a detection unit. During a single shooting task, the point light sources are sequentially illuminated, with each illumination of a point light source resulting in a single image of the object under test. Only one point light source is illuminated in each shot. Each shooting task acquires multiple images of the object under test, each image corresponding to an image illuminated by a single point light source, containing both object information and illumination information. The spatial and inter-frame features of the object images are fused to reconstruct the normal map of the object's surface. Locations in the normal map that satisfy a gradient change feature threshold are identified and marked as defects. This method can detect step defects between the polished surface and the rail base material. Fusing the spatial and inter-frame features of the object images to obtain the normal map of the object's surface removes interference from false defects caused by highlights and color differences on the polished surface, retaining only genuine processing defects in the normal map.

[0086] Optionally, the coverage areas of the light spots from point light sources within the same detection unit on the object being measured are independent of each other. This results in better reconstruction of the normal map.

[0087] Optionally, lines can be extracted from the normal diagram, and it can be determined whether the line direction is parallel to or perpendicular to the rail length direction. If a line is perpendicular to the rail length direction, it is marked as a defect. On a long rail production line, the length direction of the rail is fixed. Therefore, the length direction of the rail can be used as a judgment benchmark. In this way, transverse grinding marks on the rail can be identified.

[0088] Optionally, before obtaining the normal map of the object to be tested, the image of the object to be tested is first thresholded, and the image regions smaller than the set threshold are removed, and only the remaining regions are subjected to inter-frame feature extraction and spatial feature extraction.

[0089] The method for quantitatively measuring the machining allowance at rail welds using the aforementioned testing device or system is as follows:

[0090] A method for quantitatively measuring the surface quality of rails using laser lines and a camera. The method involves irradiating a straight light stripe onto the rail surface with a laser, capturing a rail-light stripe image with a camera, segmenting the light stripe from the image, discretely sampling points along the X-axis, extracting the center pixel coordinates of each discrete point, and identifying the light stripe center point with the smallest ordinate as a feature point. The method then searches for the minimum second derivative of the light stripe curve to the left and right of each feature point, identifying these points as rail weld boundary feature points. The difference in y-axis distance between the feature points inside and outside the rail weld boundary is used as the height of the rail bottom weld bead after grinding. The two rail weld boundary feature points represent the weld area, with the remaining area representing the rail base material.

[0091] Optionally, in some embodiments, the method for extracting the pixel coordinates of the center point of the light stripe is as follows: the vector of the normal direction at any pixel point (x0, y0) on the light stripe in the image is denoted as (n x n y The pixel coordinates of the center of the light stripe are obtained as (x0+tn). x y0+tn y ),in,

[0092] Let I(x,y) be the two-dimensional gray-level distribution function of the digital image. Thus, by utilizing the gray-level distribution characteristics of the light stripes, strong anti-interference ability and high extraction accuracy can be achieved, enabling sub-pixel level extraction of the light stripes.

[0093] Optionally, before calculating the coordinates of the center point of the light stripe, the light stripe boundary of the rail-light stripe image is coarsely extracted using the thresholding method. Then, the light stripe width w is calculated using the coarsely extracted light stripe boundary and a Gaussian convolution kernel of the corresponding width is generated. Finally, the Gaussian convolution kernel is used to calculate the center of the light stripe in the inner region of the light stripe using the above formula, thereby obtaining the coordinates of the center of the light stripe.

[0094] Let the pixel coordinates of the center of the i-th light bar be (x i y i The horizontal pixel coordinate of the center point of the light stripe is x. i The point with the smallest ordinate at the center of the light stripe is identified as the feature point C. Then, the first derivative ki is obtained by performing differential calculations and Gaussian filtering on each center point of the light stripe. Then, the filtered first derivative is differentiated and Gaussian filtered to obtain the filtered second derivative D. i , Finally, the minimum value of the second derivative is searched for on both the left and right sides from feature point C to determine the feature point of the rail weld boundary. The difference on the y-axis between the feature point inside and outside the rail weld boundary is taken as the height of the rail bottom weld bead after grinding. In this way, the machining allowance of the rail after grinding can be quantitatively measured.

[0095] Specifically, the implementation process of the light stripe center coordinate extraction method is as follows:

[0096] Step 1: Set the threshold th and input the rail-light stripe image I to be processed.

[0097] Step 2: Preprocess the rail-light stripe image I.

[0098] Step 3: Scan all columns of pixels sequentially to obtain the coordinates of the first pixel greater than th and the last pixel greater than th in that column, denoted as (B1, col) and (B2, col) respectively. These coordinates are the coordinates of the light stripe boundary obtained by coarse extraction.

[0099] Step 4: Calculate the width of the light stripe in each column using the light stripe boundaries obtained from the coarse extraction, and denote the width of the light stripe in the i-th column as w. i =|B1-B2|, then the maximum light stripe width is w max =max{w i |N≥i≥1}, where N is the number of columns in the image.

[0100] Step 5: Generate an arithmetic sequence M = {10, 20, ..., w} with a common difference of 10. max Using element m in the array i The length and width of the structure are Gaussian convolution kernel And its corresponding first-order and second-order partial derivative convolution kernels.

[0101] Step 6: For the pixels in the i-th column of the image, select the pixel in the array M generated in Step 5 that is closest to w. i element m i The constructed convolution kernel calculates the up-row coordinates of the i-th column pixel. The vector (n) of the normal direction of the pixels within the interval x n y ).

[0102] Step 7: Substitute the eigenvectors calculated in Step 5 into equation (3-21) to calculate t. When |tn x |≤0.5 and|tn y When |≤0.5, the center of the light stripe is within the current pixel. The final pixel coordinates of the light stripe center are (x0+tn) x ,y0+tn y ).

[0103] Step 8: Repeat Step 6 to Step 7 for each column, and finally the center of all light stripes can be extracted.

[0104] Specifically,

[0105] The specific algorithm steps for weld seam positioning are as follows:

[0106] Step 1: Extract the light stripe center from the line structured light image, and denote the center of the i-th light stripe as P. i The image coordinates are (x i y i ).

[0107] Step 2: Analyze the center point P of the light stripe extracted in Step 1. i Perform a traversal search and select the point with the smallest ordinate as the feature point C.

[0108] Step 3: Calculate the first derivative at each point of the light stripe center extracted in Step 1 using equation (4-2), and filter the calculation results using a one-dimensional Gaussian filter. The filter size is 51, μ is set to 0, and σ is set to 17.

[0109] Step 4: Using the first derivative obtained after Gaussian filtering in Step 3, calculate the second derivative D at each center point using the differential method. i The calculated second derivative was filtered using a one-dimensional Gaussian filter with a filter size of 51, μ set to 0, and σ set to 17.

[0110] Step 5: Search for the minimum point of the second derivative from feature point C to the left and right respectively, and determine the minimum point of the second derivative on the left and right sides of feature point C as the feature point of the rail weld boundary.

[0111] Optionally, when calculating the three-dimensional coordinates of the center of the light stripe, a system correction is performed: a horizontal auxiliary measurement reference C is introduced, and the distance m from the auxiliary measurement reference C to the calibration reference B is first determined. i Use the distance n from the measurement point to B. i Subtract the distance m from point C to point B i Once Hi is obtained, H1 and H2 at selected measurement points on both sides of the weld are measured and calculated. The misalignment amount is then calculated by |H1-H2|.

[0112] In a fourth aspect, the present invention provides a method for controlling the surface quality of the rail grinding area with high precision.

[0113] A high-precision rail grinding method uses abrasive belts, sandpaper, or grinding wheels to grind the rail surface. The method is characterized by the following steps: After grinding, defect detection is performed on the ground area. A camera and multiple point light sources configured for that camera are used as a detection unit. During a single shooting task, the point light sources are sequentially illuminated, and the camera captures an image of the object under test each time a point light source illuminates. Only one point light source is illuminated in each shot. Each shooting task acquires multiple images of the object under test. Each image corresponds to an image where a point light source illuminates the object, simultaneously containing object information and illumination information. The spatial features and inter-frame features of the object images are fused to reconstruct the normal map of the object's surface. Locations in the normal map that satisfy a gradient change feature threshold are identified and marked as defects. If a defect marker appears, an alarm is issued.

[0114] Furthermore, quantitative detection is performed before or after defect detection: A straight light stripe is irradiated onto the rail surface using a laser. The area irradiated by the light stripe includes both the ground and unground areas. A camera acquires a rail-light stripe image of the rail and the light stripe. The light stripe is segmented from the rail-light stripe image, and discrete points are taken along the X-axis. The center pixel coordinates of each discrete point are extracted, and the center point of the light stripe with the smallest ordinate is found as a feature point. From the feature point, the minimum value of the second derivative of the light stripe curve is searched to the left and right. The minimum value of the second derivative on both sides of the feature point is determined as the feature point of the rail weld boundary. The difference on the y-axis between the feature point inside and outside the rail weld boundary is used as the machining allowance before and after grinding. It is determined whether the machining allowance is within the allowance range. If it is, the quantitative requirement is considered met; otherwise, an alarm is issued.

[0115] Furthermore, defect detection and quantitative detection utilize the detection unit, detection device, or detection system described above in this invention. The method for defect detection uses the method described above in this invention. The method for quantitative detection uses the method described above in this invention.

[0116] Compared with existing technologies, this invention has the following advantages: 1. It can automatically detect qualitative defects in the surface quality of rails and quantitatively detect machining allowances, replacing manual measurement using visual inspection and micrometers on current long rail production lines, resulting in high detection efficiency and accuracy. 2. Qualitative defect detection can automatically detect milling and grinding marks at rail welds, automatically identify transverse grinding patterns and trigger alarms, and automatically identify and trigger alarms for incomplete weld removal, sharp edges, and steps. 3. It achieves quantitative measurement of rail surface machining allowances with an accuracy within 0.05mm and a measurement speed within 1 second. Attached Figure Description

[0117] Figure 1 This is a schematic diagram of the steel rail grinding equipment of the present invention.

[0118] Figure 2 This is a schematic diagram of the transverse grinding trajectory of the rail.

[0119] Figure 3 This is a schematic diagram of the limiting mechanism installed on the sanding belt device.

[0120] Figure 4 This is a schematic diagram of a grinding unit with an elastic mechanism.

[0121] Figure 5 This is a schematic diagram of two grinding units mounted on a bracket.

[0122] Figure 6 This is a schematic diagram of the polishing unit and its support, which are placed outside the polishing room.

[0123] Figure 7 This is a schematic diagram of the polishing room.

[0124] Figure 8 This is a schematic diagram showing the connection between the robotic arm and the support frame of the sanding belt device.

[0125] Figure 9 This is a schematic diagram of the longitudinal grinding structure of the present invention.

[0126] Figure 10 This is a schematic diagram of the structure of the detection device of the present invention in the first embodiment.

[0127] Figure 11 This is a partial structural schematic diagram of the detection device of the present invention.

[0128] Figure 12This is a schematic diagram of the second embodiment of the detection device of the present invention.

[0129] Figure 13 This is a schematic diagram of the structure of the second detection device of the present invention.

[0130] Figure 14 This is a schematic diagram of the upper detection component and the lower detection component of the present invention.

[0131] Figure 15 This is a schematic diagram of the structure of the detection operation component of the present invention.

[0132] Figure 16 This is a schematic diagram of the structure of the lower detection operation component of the present invention.

[0133] Figure 17 This is a schematic diagram of the network structure of the non-Lambertian surface photometric stereo model based on three-dimensional convolution of the present invention.

[0134] Figure 18 This is a flowchart illustrating the non-Lambertian surface photometric stereo method based on three-dimensional convolution of the present invention.

[0135] Figure 19 This is a diagram showing the stripe center extraction effect in a preferred embodiment of the present invention.

[0136] Figure 20 This is a schematic diagram of the accuracy verification results of the present invention.

[0137] Figure 21 The quantitative measurement method of the present invention uses a 1mm gauge block to measure the result of a single measurement. Detailed Implementation

[0138] like Figure 1-21 As shown, the steel rails mentioned in this invention refer to long profiles that are laid on railway transport lines, subways, light rails, etc., and cooperate with trains. The steel rails are I-shaped and include a top surface, a bottom surface, and a web surface located between the top surface and the bottom surface.

[0139] like Figure 1-8As shown, in some embodiments, a method for grinding a rail 10 is provided. The grinding object of this method is the rail 10 that has been online ground and welded on a long rail production line. This method uses an abrasive belt 1 to grind the rail 10 online. The method is divided into transverse grinding and longitudinal grinding. Transverse grinding refers to grinding along the height direction of the rail 10, and longitudinal grinding refers to grinding along the length direction of the rail 10. During transverse grinding, the abrasive belt 1 grinds along the first grinding trajectory 16 and the second grinding trajectory 17 from both sides of the width direction of the rail 10. The abrasive belt 1 uses flexible force to polish and remove the weld beads. The first grinding trajectory 16 and the second grinding trajectory 17 correspond to one side of the rail web. The first grinding trajectory 16 covers the top surface of the rail, and the second grinding trajectory 17 covers the bottom surface of the rail, or the first grinding trajectory 16 covers the bottom surface of the rail, and the second grinding trajectory 17 covers the top surface of the rail. During longitudinal grinding, a grinding wheel is fed longitudinally along the rail 10.

[0140] The weld bead is divided into two grinding tracks, each grinding from top to bottom, achieving continuous automated grinding of the rail top, rail web, and rail bottom. The grinding unit uses flexible force to abut against the rail 10, avoiding the problem of hard damage caused by rigid force grinding. Furthermore, if the strength of the weld bead is too great, the grinding unit will slip, preventing damage to the grinding unit itself.

[0141] like Figure 1-8 As shown, in some embodiments, during transverse grinding, the abrasive belt 1 moves not only laterally along the rail 10 but also longitudinally along the rail 10.

[0142] like Figure 1-7 As shown, in some embodiments, the sanding belt 1 is part of a grinding unit, which includes a grinding wheel 6 and a drive wheel 4. The sanding belt 1 is tensioned between the grinding wheel 6 and the drive wheel 4. The sanding belt 1, the grinding wheel 6 and the drive wheel 4 form a belt drive mechanism. During transverse grinding, the axial direction of the grinding wheel 6 is aligned with the length direction of the rail 10. The grinding wheel 6 abuts against the rail 10. The sanding belt 1 is located between the grinding wheel 6 and the rail 10. The grinding wheel 6 rotates and the sanding belt 1 removes weld beads along the tangent of the grinding wheel 6.

[0143] like Figure 1-7 As shown, in some embodiments, during transverse grinding, the force between the grinding wheel 6 and the rail 10 is monitored, and an alarm is issued when the monitored force exceeds a set force threshold.

[0144] like Figure 1 and 9 As shown, in some embodiments, longitudinal grinding includes grinding the rail web, grinding the rail top surface, and grinding the rail bottom surface. A flap wheel 71 is used to grind the rail web, a first disc grinding wheel 80 is used to grind the rail top surface, and the grinding surface of the disc angle grinding wheel is the bottom surface of the grinding wheel; a second disc grinding wheel 90 is used to grind the rail bottom surface, and the grinding surface of the disc angle grinding wheel is the bottom surface of the grinding wheel.

[0145] like Figure 1 and 9 As shown, in some embodiments, a first disc grinding wheel 80 is used to grind the curved surface transitioning from the rail web to the rail bottom, and a second disc grinding wheel 90 is used to grind the curved surface below the rail top and transitioning to the rail web; the first disc grinding wheel forms a first angle with the horizontal plane, the angle of the first angle being 0° to -10°; the second disc grinding wheel forms a second angle with the horizontal plane, the angle of the second angle being 0° to 10°.

[0146] like Figure 1 and 9 As shown, in some embodiments, the second disc grinding wheel 90 has a first grinding section and a second grinding section, which are concentric, with the first grinding section inside and the second grinding section outside. Viewed from the axial section of the grinding wheel, the second grinding section is arc-shaped, and the central angle of the second grinding section on the axial section of the grinding wheel coincides with the central angle of the curved surface below the rail top and transitioning to the rail web. This coincidence does not refer to mathematical equality, but rather that the central angle of the second grinding section is close to, but less than or equal to, the central angle of the curved surface.

[0147] like Figure 1-8 As shown, in some embodiments, the part of the grinding wheel 6 that contacts the abrasive belt 1 is a flexible component. When the grinding wheel 6 grinds the top surface of the rail, it rotates and feeds towards the center line in the width direction of the rail 10 while simultaneously moving from one side of the rail web to the other. At the same time, the grinding wheel 6 reciprocates longitudinally along the rail 10, with a lateral feed distance of not less than 35mm. When the grinding wheel 6 grinds the rail web, it rotates and feeds towards the center line in the width direction of the rail 10 while simultaneously feeding along the curved surface of the rail web in the height direction. At the same time, the grinding wheel 6 reciprocates longitudinally along the rail 10, with a lateral feed distance of not less than 35mm. This limitation on the lateral feed distance ensures that the abrasive belt 1 only grinds the weld beads without damaging the base material of the rail 10.

[0148] like Figure 1-8 As shown, in some embodiments, the lateral feed distance is 3.5 cm to 5 cm. At this lateral feed distance, weld beads are removed cleanly without damaging the base material of the rail 10.

[0149] like Figure 1-8 As shown, in some embodiments, after the welded rail 10 undergoes misalignment detection, the weld beads of the rail 10 that meet the misalignment requirements are ground using the grinding method to remove material.

[0150] like Figure 1-8 As shown, in some embodiments, before grinding, the misalignment data of the rail 10 is obtained, the higher base material on both sides of the weld of the rail 10 is found, and the grinding wheel 6 is fed laterally from the higher base material to the lower base material. In this way, the purpose of lateral grinding is achieved without damaging the base material of the rail 10.

[0151] like Figure 1-7 As shown, in some embodiments, before grinding, the height difference between the weld and the base material is measured, and the grinding wheel 6 feeds less than this height difference towards the centerline in the width direction of the rail 10. In some embodiments, this height difference is the average height difference between the weld and the base material, or the lowest height between the weld and the base material, or the median height between the weld and the base material. This avoids damage to the rail 10 base material due to overfeeding.

[0152] like Figure 1-7 As shown, in some embodiments, the belt drive mechanism formed by the sanding belt 1, the grinding wheel 6, and the drive wheel 4 is mounted on a bracket. During the grinding operation, when the sanding belt 1 remains in contact with the rail 10 for a period exceeding a set time threshold, the entire belt drive mechanism stops feeding towards the rail 10. The bracket then drives the grinding wheel 6 to retract away from the rail 10 until the force between the grinding wheel 6 and the rail 10 is reduced to an acceptable range. This prevents damage or failure of the belt drive mechanism.

[0153] One scheme for the support to drive the grinding wheel 6 to retract away from the rail 10 is as follows: the support part for mounting the grinding wheel 6 and the support part for mounting the transmission wheel 4 are independent of each other, and the two support parts are connected together by a flexible or elastic component. When the time the sanding belt 1 is in contact with the rail 10 does not exceed the set time threshold, the flexible or elastic component keeps the relative position of the grinding wheel 6 and the transmission wheel 4 stable, and the grinding wheel 6 performs the grinding operation; when the time the sanding belt 1 is in contact with the rail 10 exceeds the set time threshold, the flexible or elastic component deforms, causing the grinding wheel 6 to move closer to the transmission wheel 4.

[0154] Another scheme involves the support retracting the grinding wheel 6 away from the rail 10. The support for the grinding wheel 6 is mounted on a frame. An elastic or flexible element is installed between the support and the frame to create a flexible force during grinding. When the contact time between the sanding belt 1 and the rail 10 does not exceed a set time threshold, the flexible or elastic element stabilizes the relative position of the support and the frame, allowing the grinding wheel 6 to perform the grinding operation. When the contact time exceeds the set time threshold, the flexible or elastic element deforms, causing the support to move closer to the frame. The frame can be connected to a mechanized or automated feed power unit to feed the grinding wheel 6.

[0155] These two types of brackets, which drive the grinding wheel 6 to retract away from the rail 10, can be used independently or combined to form a composite flexible force mechanism.

[0156] like Figure 1-8 As shown, in some embodiments, a rail 10 grinding system includes a grinding unit, a rail 10 support mechanism, and a feeding mechanism for feeding the grinding unit. The feeding mechanism is connected to the grinding unit, and a buffer assembly is provided between the grinding unit and the feeding mechanism. The buffer assembly includes a spring or a flexible element other than a spring. The grinding unit includes a grinding wheel 6, a transmission wheel 4, and a sanding belt 1. The sanding belt 1 is tensioned to the grinding wheel 6 and the transmission wheel 4. The sanding belt 1, the grinding wheel 6, and the transmission wheel 4 form a belt drive mechanism. During transverse grinding, the axial direction of the grinding wheel 6 is aligned with the length direction of the rail 10.

[0157] like Figure 1-8 As shown, in some embodiments, the grinding unit includes a mounting frame, a grinding wheel 6 and a drive wheel 4 are mounted on the mounting frame, a tensioning wheel 5 is provided on the mounting frame, the tensioning wheel 5 is connected to the tensioning drive 3, and the tensioning wheel 5 tensions the sanding belt 1 to the grinding wheel 6 and the drive wheel 4; when performing a grinding task, the grinding wheel 6 abuts against the rail 10; the grinding unit has a damping mechanism that limits the force threshold of the grinding wheel 6, when the force between the grinding wheel 6 and the rail 10 is greater than the force threshold, the grinding wheel 6 is displaced towards the drive wheel 4.

[0158] like Figure 8 As shown, in some embodiments, the grinding wheel 6 includes a rotating shaft and a flexible sleeve 8, with the rotating shaft inside and the flexible sleeve 8 outside, and the sanding belt 1 in contact with the flexible sleeve 8.

[0159] like Figure 1-8 As shown, in some embodiments, the mounting frame is provided with a tensioning drive 3 and a tensioning wheel 5. The tensioning drive 3 includes a fixed part 9 and a movable part 7. The tensioning wheel 5 is disposed on the movable part 7, and the fixed part 9 is disposed on the mounting frame.

[0160] like Figure 1-8 As shown, in some embodiments, the mounting bracket is provided with a limiting roller, the axis of which is parallel to the axis of the grinding wheel 6; when the limiting roller touches the rail 10, the grinding wheel 6 reaches the limit position of feeding into the rail 10.

[0161] like Figure 1-8 As shown, in some embodiments, the mounting bracket includes a fixed part 9 and a movable part 7, with the grinding wheel 6 located in the movable part 7 and the transmission wheel 4 located in the fixed part 9; when the force between the grinding wheel 6 and the rail 10 is greater than the force threshold, the movable part 7 retracts; when the force between the grinding wheel 6 and the rail 10 is less than the force threshold, the relative positions of the movable part 7 and the fixed part 9 are fixed.

[0162] like Figure 1-8As shown, in some embodiments, the mounting bracket is provided with a damping mechanism, which includes a spring and a locking member. The spring is disposed between the movable part 7 and the fixed part 9. The spring and the locking member work together to keep the movable part 7 in a balanced position, and the distance between the grinding wheel 6 and the transmission wheel 4 is fixed.

[0163] like Figure 4 As shown, in some embodiments, the fixed part 9 is provided with a first spring mounting member 12, and the movable part 7 is provided with a second spring mounting member 13, with the two ends of the spring being fixed to the first spring mounting member 12 and the second spring mounting member 13 respectively.

[0164] like Figure 5-6 As shown, in some embodiments, the center of the grinding wheel 6 and the center of the drive wheel 4 are at opposite heights, with the tension wheel 5 located between the grinding wheel 6 and the drive wheel 4. The triangle formed by the centers of the grinding wheel 6, the drive wheel 4, and the tension wheel 5 lies within the area enclosed by the sanding belt 1. This arrangement provides good tension stability for the sanding belt 1, and the friction between the sanding belt 1 and the grinding wheel 6 is adjustable, which helps maintain a stable relative position between the sanding belt 1 and the grinding wheel 6 when the grinding wheel 6 rotates at high speed.

[0165] like Figure 5 As shown, in some embodiments, the tensioning drive 3 is a cylinder, hydraulic cylinder or electric push rod, and the tensioning drive 3 intersects the straight line formed by the center of the grinding wheel 6 and the center of the transmission wheel 4. The tensioning wheel 5 is closer to the transmission wheel 4 than the grinding wheel 6.

[0166] like Figure 1-7 As shown, in some embodiments, when the sanding belt 1 is tensioned, the section of sanding belt 1 between the grinding wheel 6 and the drive wheel 4 forms an angle with the horizontal plane, the section of sanding belt 1 between the grinding wheel 6 and the tension wheel 5 forms an angle with the horizontal plane, the sections of sanding belt 1 between the grinding wheel 6 and the drive wheel 4 and the section of sanding belt 1 between the grinding wheel 6 and the tension wheel 5 form an acute angle, and the section of sanding belt 1 between the drive wheel 4 and the tension wheel 5 forms an obtuse angle with the section of sanding belt 1 between the grinding wheel 6 and the tension wheel 5. This maintains the stability of the grinding unit and the sanding belt 1.

[0167] In some embodiments, the acute angle is less than 45°, and the obtuse angle is less than 150° and greater than 120°.

[0168] like Figure 1-7As shown, in some embodiments, a first reference plane is constructed, which passes through the axle of the grinding wheel 6 and the axle of the transmission wheel 4. The center plane of the grinding wheel 6 in the width direction is used as a second reference plane, which is perpendicular to the first reference plane. The intersection of the second reference plane and the axle of the tension wheel 5 is projected onto the line of intersection of the first and second reference planes to obtain the projection point of the tension wheel 5. The point of intersection of the axle of the grinding wheel 6 and the first and second reference planes is used as the projection point of the grinding wheel 6. The point of intersection of the axle of the transmission wheel 4 and the first and second reference planes is used as the projection point of the transmission wheel 4. The ratio of the distance between the projection point of the tension wheel 5 and the projection point of the grinding wheel 6 to the distance between the projection point of the tension wheel 5 and the projection point of the transmission wheel 4 is at least 5:4.

[0169] As shown in 6-7, in some embodiments, the rail 10 grinding system has two grinding units as described above, which are respectively arranged on both sides of the rail 10 in the width direction, with one grinding unit higher and the other grinding unit lower.

[0170] like Figure 9 As shown, in some embodiments, the tension wheel 5 of one grinding unit faces upward and the tension wheel 5 of the other grinding unit faces downward.

[0171] like Figure 1 and 5 As shown, in some embodiments, the polishing system includes a bracket for mounting polishing units, with two polishing units mounted on the same bracket; the bracket includes a horizontal arm 330 and two connecting arms located on both sides of the horizontal arm 330, with a polishing unit mounted on each connecting arm, and an elastic mechanism provided between the connecting arm and the horizontal arm 330; the horizontal arm 330 is provided with an electrical connector 332 for connecting to the robot.

[0172] like Figure 1 and 5 As shown, in some embodiments, the ratio of the distance from the center of the grinding wheel 6 of the high grinding unit to the horizontal arm 330 to the distance from the center of the grinding wheel 6 of the low grinding unit to the horizontal arm 330 is 0.7 to 0.9. With this configuration, when switching from grinding one side of the rail to the other side, the pose of the robotic arm or robot itself remains basically unchanged; only the pose of the support needs to be switched.

[0173] like Figure 5As shown, in some embodiments, the connecting arm includes a slant plate 3310 connected to the cross arm 330 and a tripod 3311 connected to the slant plate 3310. The mounting bracket of the grinding unit is connected to the slant plate 3310 and the tripod 3311. The mounting bracket is located at the bottom of the slant plate 3310 and the tripod 3311, forming a triangular structure. The center of the transmission wheel 4 is located within the area covered by the triangular structure, and the grinding wheel 6 is located outside the triangular structure. A flexible mechanism can be provided between the slant plate 3310 and the cross arm 330, such as a damper, allowing the slant plate 3310 and the cross arm 330 to have millimeter-level rotational displacement. The damper can be a spring, a rubber pad, a toothed connector, etc. A flexible mechanism can also be provided between the tripod 3311 and the slant plate 3310. The rigidity (or flexibility) of the connection between the tripod 3311 and the inclined plate 3310 can be different from the rigidity (or flexibility) level or degree of the connection between the inclined plate 3310 and the cross arm 330, so that the entire support has multiple flexibility mechanisms.

[0174] like Figure 5 As shown, in some embodiments, the movable part 7 of the mounting bracket is located outside the triangular structure, and the fixed part 9 is part of the triangular structure; the mounting bracket is provided with a limiting block, and the triangular bracket 3311 blocks and limits the limiting block.

[0175] like Figure 8 As shown, in some embodiments, the connecting arm is provided with a cooling medium pipe and a cooling control valve. The cooling medium pipe extends from the inclined plate 3310 toward the tripod 3311, and the tripod 3311 is provided with a positioning member 142 for the cooling medium pipe. The cooling medium pipe includes a rigid pipe 140 and a flexible pipe 141. The rigid pipe 140 is connected to the positioning member 142, and the flexible pipe 141 is connected to the cooling control valve and the rigid pipe 140.

[0176] like Figure 5 As shown, in some embodiments, the feeding mechanism is a robotic arm, and a quick-change interface is provided on the horizontal arm 330. The connection surface between the quick-change interface and the robotic arm is inclined relative to the horizontal arm 330.

[0177] In some embodiments, the feeding mechanism is a three-dimensional moving platform, with the center of the cross arm 330 connected to the three-dimensional moving platform. The three-degree-of-freedom moving platform provides translational displacement of the support along the X, Y, and Z axes.

[0178] In some embodiments, the feeding mechanism is a three-dimensional moving platform, which is connected to the mounting frame of each grinding unit. In this case, the grinding unit is directly connected to the three-dimensional moving platform without the need for a support structure. For example, the three-dimensional moving platform uses a plane parallel to the bottom of the rail as the XOY plane, the length direction of the rail 10 as the X-axis, and the height direction of the rail 10 as the Z-axis. The three-dimensional moving platform enables the sanding belt 1 mechanism to translate along the x-axis, y-axis, and z-axis to achieve grinding feed; each grinding unit is mounted on its own Z-axis translation module, and the two grinding units share the x-axis and y-axis translation modules.

[0179] like Figure 6-7 As shown, in some embodiments, the aforementioned rail 10 grinding system is installed inside a grinding chamber, which isolates the grinding work space from the external environment, thus enabling unmanned grinding operations.

[0180] In some embodiments, the grinding chamber has a rotating wall that can rotate around its centerline. There is a sealing mechanism between the boundary of the rotating wall and the grinding chamber. The rotating wall is set on a turntable, the center of which is connected to a turntable motor. A pair of shelves are provided on the turntable, with the two shelves located on opposite sides of the rotating wall.

[0181] In this way, it is possible to replace worn parts such as sanding belts and grinding wheels or perform maintenance outside the grinding chamber without stopping the machine inside.

[0182] like Figure 5 As shown, in some embodiments, the tensioning drive 3 is a tensioning cylinder, the shelf is provided with an air supply port, and the sanding belt 1 mechanism is provided with an air inlet. When the support of the grinding unit is placed on the shelf, the air supply port is connected to the air inlet. The air supply port is connected to the air source. The shelf is provided with a force sensor. When the force is sensed, the air source is turned on. When the force is 0, the air source is turned off.

[0183] The aforementioned rail 10 grinding system is applied to the long rail production line. It uses polishing and grinding to remove weld beads on the rail 10 welds, forming a rail 10 rough grinding workstation for the long rail production line.

[0184] In some embodiments, a rough grinding workstation for rail 10 is used in a long rail production line. The rail 10 grinding system is installed on the long rail production line after the misalignment detection station after welding.

[0185] This invention provides an automated rail inspection system for detecting the surface quality of rails. For example, it performs quantitative and qualitative testing on the surface quality of polished rails.

[0186] like Figure 10-16As shown, in some embodiments, an automated detection device includes a camera 403, a point light source 402, and a detection unit support 440. The camera 403 and the point light source 402 are mounted on the detection unit support 440. A single camera 403 and multiple point light sources 402 configured for that camera 403 constitute a detection unit. The coverage areas of the light spots of the point light sources 402 within the same detection unit on the object under test are independent. During a single shooting task, the point light sources 402 are sequentially illuminated. Each time a point light source 402 illuminates, the camera 403 captures an image of the object under test. Only one point light source 402 is illuminated during each shot. During a single shooting task, the positions of the point light sources 402 and the camera 403 are relatively fixed. Each shooting task acquires multiple images of the object under test. Each image corresponds to an image where a point light source 402 illuminates the object, simultaneously containing object information and illumination information. The spatial features and inter-frame features of the object images are fused to obtain a three-dimensional image of the surface texture of the object under test.

[0187] In some embodiments, spatial features and inter-frame features are extracted from the image of the object under test. Both spatial features and inter-frame features are represented by three-dimensional convolutions. The three-dimensional convolution includes a two-dimensional spatial dimension and a one-dimensional inter-frame dimension. The value of the one-dimensional inter-frame dimension in the three-dimensional convolution of spatial features is a set value, and the two-dimensional spatial dimension is the spatial feature value of the image. The value of the two-dimensional spatial dimension in the three-dimensional convolution of inter-frame features is a set value, and the one-dimensional inter-frame dimension is the inter-frame feature value. When performing image information fusion, the inter-frame features are processed first, then the spatial features are processed, and then the inter-frame features and spatial features are fused to obtain a three-dimensional image of the surface texture of the object under test.

[0188] like Figure 11 As shown, in some embodiments, the point light source 402 is fixed to the detection unit bracket 440 by the light source bracket 50. Each light source bracket 50 is provided with a point light source 402. The light source bracket 50 includes a hanging ear 500 and a hanging piece 501. The hanging ear 500 is fixed to the detection unit bracket 440, and the light source is fixed on the hanging piece 501. The hanging piece 501 and the hanging ear 500 are rotatably connected, and there is damping between the hanging piece 501 and the hanging ear 500.

[0189] like Figure 11 As shown, in some embodiments, the ear loop 500 has an ear portion 5000 extending away from the detection unit support 440, and the pendant 501 has a beam and a connecting portion 5010 extending towards the detection unit support 440. The point light source 402 is fixed to the beam. The connecting portion 5010 overlaps with the ear portion 5000, and a light source pivot is located between the connecting portion 5010 and the ear portion 5000. The light source pivot is fixed to the connecting portion 5010 or the ear portion 5000. Rotating the beam allows adjustment of the illumination area of ​​the light source on the object being measured.

[0190] like Figure 11 As shown, in some embodiments, the light source shaft is fixed to the connecting part 5010, and the light source shaft is connected to the output shaft of the light source drive motor. The presence of the light source shaft allows the position of the light source to be adjusted, thereby adapting to objects of different sizes and specifications. Using an encoder or angle measuring instrument, the relationship between the size of the object being measured and the rotation angle of the light source can be obtained, facilitating rapid calibration of the light source position when changing the object being measured.

[0191] In some embodiments, with camera 403 as the center and any plane perpendicular to the axis of camera 403 as the reference plane, the light sources are grouped according to their distance from camera 403. Light sources in the same group have equal distances from camera 403, while light sources in different groups have random distances from camera 403. There are at least 3 light sources in the same group.

[0192] In some embodiments, the polar coordinate angle of the light source is randomized with the camera 403 as the center. In this way, the light source can illuminate the object being measured as discretely as possible.

[0193] In some embodiments, the ratio of the distance between any two adjacent light sources to the shooting distance from the camera 403 to the object under test is 0.1 to 0.2. With this setting, the image captured by the camera 403 is clear, and the illumination areas of the light sources on the object under test do not overlap, resulting in a good 3D image reconstruction effect of the surface texture of the object under test.

[0194] During the production of long rails, the surface quality requirements for rail 10 include: the weld and the weld slag around the weld must be cleaned, the ground surface must be smooth, the rail base (the curved part that transitions from the rail web to the rail base) must be rounded, and it is not allowed to grind laterally or damage the rail 10 base material.

[0195] When the aforementioned automated inspection device is applied to the surface quality inspection of rails, an automated rail inspection system is formed. In some embodiments, an automated rail 10 inspection system includes at least one set of the aforementioned inspection devices. The inspection object of the system is a rail 10 located on the production line that has completed the initial milling or initial grinding of weld beads. The inspection system is set after the first removal process of the weld of the rail 10, or there are two sets of inspection systems: the first set of inspection systems is set after the detection of errors in the rail 10 and before the first removal process of the weld of the rail 10; the second set of inspection systems is set after the first removal process of the weld of the rail 10.

[0196] like Figure 12As shown, in some embodiments, the rail 10 detection system has a frame 60, an automated detection device is installed inside the frame 60, the frame 60 has side plates and a top plate, the side plates and the top plate enclose the interior of the frame 60 into a dark chamber, one of the side plates has an entrance that allows the rail 10 to enter, and the opposite side plate has an exit that allows the rail 10 to exit.

[0197] like Figure 12 As shown, in some embodiments, the side panel has a door or bed, or the side panel is detachably connected to the frame 60.

[0198] In the production of long rails, the surface inspection requirements for rail 10 also include: conducting full-section inspection of the weld seams of rail 10.

[0199] like Figure 12 As shown, in some embodiments, the frame 60 is provided with a detection drive device that can move 360° around the cross section of the rail 10. The detection device is connected to the detection drive device through an interface, and the detection drive device carries the detection device to perform detection operations on each surface of the rail 10 in a step-by-step manner.

[0200] In some embodiments, there is only one detection device, and the detection drive device carries the detection device in a step-by-step manner to perform detection operations on the top surface of the rail, the web surface of the rail, the bottom surface of the rail, and another web surface of the rail.

[0201] like Figure 12-16 As shown, in some embodiments, the frame 60 has an upper detection operation assembly and a lower detection operation assembly 62. The upper detection operation assembly has a first detection device 610 parallel to the ground, a second detection device 611 and a third detection device 612 located on both sides of the first detection device 610, and the second detection device 611 and the third detection device 612 are centrally symmetrical about the camera 403 of the first detection device 610.

[0202] like Figure 13-14 As shown, in some embodiments, the upper detection operation component is connected to the detection drive device. The detection object of the second detection device 611 is a rail web surface, and the third detection device 612 is another rail web surface. The second detection device 611 and the third detection device 612 each have three detection positions. The detection drive device causes the second detection device 611 or the third detection device 612 to migrate at the three detection positions respectively. At the first detection position, the angle between the axis of the camera 403 of the second detection device 611 and the horizontal plane is -25° to -38°. At the second detection position, the axis of the second detection device 611 is parallel to the horizontal plane. At the third detection position, the angle between the axis of the second detection device 611 and the horizontal plane is 35° to 55°.

[0203] like Figure 12 and 16As shown, in some embodiments, when the second detection device 611 is operating, the third detection device 612 is not operating; when the third detection device 612 is operating, the second detection device 611 is not operating.

[0204] like Figure 12 and 16 As shown, in some embodiments, the lower detection operation component 62 is fixed below the rail base.

[0205] In some embodiments, the detection device comprises four sets, each set corresponding to a detection area of ​​one rail 10. The detection devices surround the detection area, and the rail 10 enters the detection area along the axial direction of the prism. The distance between two opposing detection devices is not less than 1050 mm. This enables the detection device to be applicable to the detection of rails 10 on long rail production lines.

[0206] like Figure 13-14 As shown, in some embodiments, one set of detection devices is aligned with the top surface of the rail, one set of detection devices is aligned with the bottom surface of the rail, one set of detection devices is aligned with one rail web surface, and another set of detection devices is aligned with another rail web surface. Any set of detection devices on the rail web surface has an upper detection unit 64, a middle detection unit 65, and a lower detection unit 66. Each detection unit has a camera 403, a point light source 402, and a light source controller. The angle between the axis of the camera 403 of the upper detection unit 64 and the horizontal plane is 27° to 34° downward, the axis of the camera 403 of the middle detection unit 65 is parallel to the horizontal plane, and the angle between the axis of the camera 403 of the lower detection unit 66 and the horizontal plane is 41° to 49° upward.

[0207] like Figure 13-14 As shown, in some embodiments, any set of detection devices on the rail web includes a back plate. The upper detection unit 64, the middle detection unit 65, and the lower detection unit 66 are respectively installed on the back plate through their respective position adjustment mechanisms. The back plate includes a main plate 630, an upper wing plate 631, and a lower wing plate 632. The main plate 630 is parallel to the height direction of the rail 10 to be tested. The upper wing plate 631 intersects with the main plate 630, and the top of the upper wing plate 631 is closer to the rail 10 to be tested than the bottom of the upper wing plate 631. The lower wing plate 632 intersects with the main plate 630, and the top of the lower wing plate 632 is farther away from the rail 10 to be tested than the bottom of the lower wing plate 632.

[0208] like Figure 13 As shown, in some embodiments, the camera 403 of the upper detection unit 64 is located near the area where the upper wing plate 631 and the main board 630 intersect, a portion of the light source of the upper detection unit 64 is located in the area covered by the upper wing plate 631, and another portion of the light source of the upper detection unit 64 is located in the area covered by the main board 630.

[0209] like Figure 13As shown, in some embodiments, the camera 403 of the detection unit 65 and its light source are disposed in the area covered by the motherboard 630.

[0210] like Figure 13 As shown, in some embodiments, the camera 403 of the lower detection unit 66 and its light source are disposed in the area covered by the lower wing plate 632.

[0211] In this way, any set of detection devices on the rail web surface can perform full surface detection of the surface of the rail web surface located below the rail top, the side surface of the rail web, the small arc surface transitioning between the surface below the rail top and the side surface of the rail web, the large arc surface transitioning from the rail web to the rail bottom, and the vertical surface adjacent to the rail web and the rail bottom.

[0212] In the production of long rails, the surface inspection requirements for rail 10 also include: the height of the bottom weld bead (weld) after grinding is not less than 0.5mm. Therefore, it is also necessary to perform quantitative measurement on the ground rail 10.

[0213] like Figure 17 As shown, an optional scheme for the quantitative measurement device is: the detection device has a line laser 67, the axis of the line laser 67 is at an angle to the axis of the camera 403, the line laser 67 does not emit laser when the light source is lit; the light source is not lit when the line laser 67 emits laser lines.

[0214] Based on the surface quality requirements of rail 10, we specify the surface defect detection indicators of the ground rail 10 as: whether there is a transverse grinding track, whether there is a step between the grinding surface and the rail 10 base material, and propose a method for detecting surface defects of rail 10 using the above-mentioned detection device or detection system.

[0215] In some embodiments, a method for detecting surface defects of rail 10 using a group of cameras 403 and point light sources 402 is provided. This method uses one camera 403 and multiple point light sources 402 configured for that camera 403 as a detection unit. During a single shooting task, the point light sources 402 are sequentially illuminated. Each time a point light source 402 illuminates, the camera 403 captures an image of the object under test. Only one point light source 402 is illuminated during each shot. Each shooting task acquires multiple images of the object under test. Each image corresponds to an image where a point light source 402 illuminates the object, simultaneously containing object information and illumination information. The spatial features and inter-frame features of the object images are fused to reconstruct the normal map of the object's surface. Locations in the normal map that satisfy a gradient change feature threshold are identified and marked as defects. In this way, step defects between the polished surface and the rail 10 base material can be detected. By fusing spatial and inter-frame features of the image of the object under test, a normal map of the object's surface can be obtained. This can remove interference from false defects caused by highlights and color differences on the polished surface, and retain only the true processing defects in the normal map.

[0216] In some embodiments, the coverage areas of the light spots of the point light source 402 within the same detection unit on the object under test are independent of each other. This results in good reconstruction of the normal map.

[0217] In some embodiments, lines are extracted from the normal diagram, and it is determined whether the line direction is parallel to or perpendicular to the length direction of the rail 10. If a line is perpendicular to the length direction of the rail 10, it is marked as a defect. On a long rail production line, the length direction of the rail 10 is fixed. Therefore, the length direction of the rail 10 can be used as a judgment criterion. In this way, transverse grinding marks on the rail 10 can be identified.

[0218] like Figure 15-18 As shown, in some embodiments, before obtaining the normal map of the object to be tested, the image of the object to be tested is first thresholded, and the image regions smaller than the set threshold are removed, and only the remaining regions are subjected to inter-frame feature extraction and spatial feature extraction.

[0219] This invention discloses a non-Lambertian surface photometric stereo model based on three-dimensional convolution, including an information fusion layer 100, inter-frame feature extraction 101, spatial feature extraction 102, a max pooling layer 103, and a regression layer 104.

[0220] The method for quantitatively measuring the machining allowance at rail welds using the aforementioned testing device or system is as follows:

[0221] A method for quantitatively measuring the surface quality of rails using laser lines and a camera. The method involves irradiating the rail surface with a straight light stripe using a laser (67), acquiring a rail-light stripe image using a camera, segmenting the light stripe from the image, discretely sampling points along the X-axis, extracting the center pixel coordinates of each discrete point, and identifying the light stripe center point with the smallest ordinate as a feature point. The method then searches for the minimum second derivative of the light stripe curve to the left and right of each feature point, identifying these points as rail weld boundary feature points. The difference in y-axis distance between the feature points inside and outside the rail weld boundary is used as the height of the rail bottom weld bead after grinding. The two rail weld boundary feature points represent the weld area, with the remaining area representing the rail base material.

[0222] Optionally, in some embodiments, the method for extracting the pixel coordinates of the center point of the light stripe is as follows: the vector of the normal direction at any pixel point (x0, y0) on the light stripe in the image is denoted as (n x n y The pixel coordinates of the center of the light stripe are obtained as (x0+tn). x y0+tn y ),in,

[0223] Let I(x,y) be the two-dimensional gray-level distribution function of the digital image. Thus, by utilizing the gray-level distribution characteristics of the light stripes, strong anti-interference ability and high extraction accuracy can be achieved, enabling sub-pixel level extraction of the light stripes.

[0224] Optionally, before calculating the coordinates of the center point of the light stripe, the light stripe boundary of the rail-light stripe image is coarsely extracted using the thresholding method. Then, the light stripe width w is calculated using the coarsely extracted light stripe boundary and a Gaussian convolution kernel of the corresponding width is generated. Finally, the Gaussian convolution kernel is used to calculate the center of the light stripe in the inner region of the light stripe using the above formula, thereby obtaining the coordinates of the center of the light stripe.

[0225] Let the pixel coordinates of the center of the i-th light bar be (x i y i The horizontal pixel coordinate of the center point of the light stripe is x. i The point with the smallest ordinate at the center of the light stripe is identified as the feature point C. Then, the first derivative ki is obtained by performing differential calculations and Gaussian filtering on each center point of the light stripe. Then, the filtered first derivative is differentiated and Gaussian filtered to obtain the filtered second derivative D. i , Finally, the minimum value of the second derivative is searched for on both the left and right sides from feature point C to determine the feature point of the rail weld boundary. The difference on the y-axis between the feature point inside and outside the rail weld boundary is taken as the height of the rail bottom weld bead after grinding. In this way, the machining allowance of the rail after grinding can be quantitatively measured.

[0226] Specifically, the implementation process of the light stripe center coordinate extraction method is as follows:

[0227] Step 1: Set the threshold th and input the rail-light stripe image I to be processed.

[0228] Step 2: Preprocess the rail-light stripe image I.

[0229] Step 3: Scan all columns of pixels sequentially to obtain the coordinates of the first pixel greater than th and the last pixel greater than th in that column, denoted as (B1, col) and (B2, col) respectively. These coordinates are the coordinates of the light stripe boundary obtained by coarse extraction.

[0230] Step 4: Calculate the width of the light stripe in each column using the light stripe boundaries obtained from the coarse extraction, and denote the width of the light stripe in the i-th column as w. i =|B1-B2|, then the maximum light stripe width is w max =max{w i |N≥i≥1}, where N is the number of columns in the image.

[0231] Step 5: Generate an arithmetic sequence M = {10, 20, ..., w} with a common difference of 10. max Using element m in the array i The length and width of the structure are Gaussian convolution kernel And its corresponding first-order and second-order partial derivative convolution kernels.

[0232] Step 6: For the pixels in the i-th column of the image, select the pixel in the array M generated in Step 5 that is closest to w. i element m i The constructed convolution kernel calculates the up-row coordinates of the i-th column pixel. The vector (n) of the normal direction of the pixels within the interval x n y ).

[0233] Step 7: Substitute the eigenvectors calculated in Step 5 into equation (3-21) to calculate t. When |tn x |≤0.5 and|tn y When |≤0.5, the center of the light stripe is within the current pixel. The final pixel coordinates of the light stripe center are (x0+tn) x ,y0+tny ).

[0234] Step 8: Repeat Step 6 to Step 7 for each column, and finally the center of all light stripes can be extracted.

[0235] Specifically,

[0236] The specific algorithm steps for weld seam positioning are as follows:

[0237] Step 1: Extract the light stripe center from the line structured light image, and denote the center of the i-th light stripe as P. i The image coordinates are (x i y i ).

[0238] Step 2: Analyze the center point P of the light stripe extracted in Step 1. i Perform a traversal search and select the point with the smallest ordinate as the feature point C.

[0239] Step 3: Calculate the first derivative at each point of the light stripe center extracted in Step 1 using equation (4-2), and filter the calculation results using a one-dimensional Gaussian filter. The filter size is 51, μ is set to 0, and σ is set to 17.

[0240] Step 4: Using the first derivative obtained after Gaussian filtering in Step 3, calculate the second derivative D at each center point using the differential method. i The calculated second derivative was filtered using a one-dimensional Gaussian filter with a filter size of 51, μ set to 0, and σ set to 17.

[0241] Step 5: Search for the minimum point of the second derivative from feature point C to the left and right respectively, and determine the minimum point of the second derivative on the left and right sides of feature point C as the feature point of the rail weld boundary.

[0242] Optionally, when calculating the three-dimensional coordinates of the center of the light stripe, a system correction is performed: a horizontal auxiliary measurement reference C is introduced, and the distance m from the auxiliary measurement reference C to the calibration reference B is first determined. i Use the distance n from the measurement point to B. i Subtract the distance m from point C to point B i Hi is obtained, and H1 and H2 are calculated at selected measurement points on both sides of the weld. The misalignment is then calculated using |H1-H2|. Using the above quantitative measurement method, the light stripe center of the object image is extracted and converted to coordinates in the world coordinate system. Since the line structured light is obliquely incident on the gauge block during measurement, the point cloud of the gauge block is tilted. To obtain the thickness measurement value of the gauge block, it is only necessary to take the point cloud at the Z... wThe coordinates along the axis are sufficient to obtain the final measurement result of the 1mm gauge block, as shown below. Figure 21 As shown.

[0243] The measurement was repeated 10 times, and the final results are shown in Table 3.4.

[0244] Table 3.4 Measurement results of 1mm gauge block

[0245]

[0246] As shown in Table 3.4, the mean measurement of the 1mm gauge block is 1.0161mm, the mean deviation is 0.0161mm, the standard deviation of the measurement is 0.0083mm, and the range is 0.0255mm. Therefore, the measurement accuracy of this measurement system meets the measurement requirements.

[0247] In a fourth aspect, the present invention provides a method for controlling the surface quality of the rail grinding area with high precision.

[0248] In some embodiments, a high-precision rail grinding method uses abrasive belts, sandpaper, or grinding wheels to grind the rail surface. After grinding, defect detection is performed on the ground area: a camera 403 and multiple point light sources 402 configured for the camera are used as a detection unit. During a single shooting task, the point light sources are lit sequentially. Each time a point light source is lit, the camera captures an image of the object under test. Each time the camera captures an image, only one point light source is lit. Each shooting task acquires multiple images of the object under test. Each image of the object under test corresponds to an image with both object information and illumination information when a point light source illuminates the object under test. The spatial features and inter-frame features of the images of the object under test are fused to reconstruct the normal map of the surface of the object under test. The location in the normal map that satisfies the gradient change feature threshold is found and marked as a defect. If a defect mark is found, an alarm is issued.

[0249] In some embodiments, quantitative detection is performed before or after defect detection: a straight light stripe is irradiated onto the rail surface by a laser 67. The area irradiated by the light stripe includes the ground area and the unground area. A camera acquires a rail-light stripe image of the rail and the light stripe. The light stripe is segmented from the rail-light stripe image. Points are discretely selected along the X-axis of the light stripe, and the center pixel coordinates of each discrete point are extracted. The center point of the light stripe with the smallest ordinate is found as a feature point. The minimum value of the second derivative of the light stripe curve is searched for on both the left and right sides of the feature point. The minimum value of the second derivative on both sides of the feature point is determined as the feature point of the rail weld boundary. The difference on the y-axis between the feature point inside the rail weld boundary and the feature point outside the rail weld boundary is used as the processing allowance after grinding and before grinding. It is determined whether the processing allowance is within the allowance range. If so, the quantitative requirement is considered to be met; otherwise, an alarm is issued.

[0250] In some embodiments, defect detection and quantitative detection utilize the detection unit, detection device, or detection system described above in this invention. The method for defect detection uses the method described above in this invention. The method for quantitative detection uses the method described above in this invention.

[0251] The various embodiments of the present invention can be used independently as technical solutions, or they can be combined to form a combined solution.

[0252] It is understood that the embodiments described in this invention are preferred embodiments and features, and any person skilled in the art can make some modifications and variations based on the spirit of the invention described herein. Such modifications and variations are also considered to fall within the scope of this invention and the limits set forth in the independent and appended claims.

Claims

1. A rail weld grinding system, comprising a grinding chamber and a control system, wherein the grinding chamber is equipped with a grinding system and a vision positioning system, characterized in that: The grinding system includes a robotic arm and an end effector. The end effector includes a sanding belt mechanism and a support. The support includes a horizontal arm and two connecting arms located on both sides of the horizontal arm. Each connecting arm is equipped with a sanding belt mechanism, and the two sanding belt mechanisms are opposite each other. The sanding belt mechanism includes a mounting frame, a grinding wheel, a drive wheel, a tension wheel, and a sanding belt mounted on the mounting frame. The sanding belt, grinding wheel, and drive wheel form a belt drive mechanism. The tension wheel is connected to a tensioning drive component. During the grinding task, the grinding wheel rests against the rail. The sanding belt mechanism has a damping mechanism that limits the force threshold of the grinding wheel. When the force between the grinding wheel and the rail exceeds the force threshold, the grinding wheel moves towards the drive wheel. The vision positioning system includes a camera; the end effector and the robotic arm are connected via a quick interface; the control system has a pre-set grinding trajectory, which is divided into a first trajectory and a second trajectory, forming a complete circle of weld seam; each sanding belt mechanism corresponds to one trajectory; The visual positioning system also includes a point light source and a detection unit support; the camera and the point light source are set on the detection unit support; with the camera as the center and any plane perpendicular to the camera axis as the reference plane, the light sources are grouped according to their distance from the camera. The distance between the light sources in the same group and the camera is equal, while the distance between the light sources in different groups and the camera is random. The rail weld grinding room has a door, and the grinding system is set inside the grinding room. The grinding room is equipped with a rotating wall that can rotate around its centerline. A sealing mechanism is provided between the boundary of the rotating wall and the grinding room. The rotating wall is set on a turntable, and the center of the turntable is connected to the turntable motor. A pair of shelves are provided on the turntable, and the two shelves are located on both sides of the rotating wall. The tensioning wheel of the sanding belt mechanism is driven by a tensioning cylinder. The shelf is equipped with an air inlet, and the sanding belt mechanism is equipped with an air inlet. When the sanding belt mechanism is placed on the shelf, the air inlet and the air inlet are connected. The air inlet is connected to an air source. The shelf is equipped with a sensor that senses when the sanding belt mechanism is placed on the shelf. When the sensor senses that the sanding belt mechanism is placed on the shelf, the air source is turned on. When the sensor senses that the sanding belt mechanism is removed from the shelf, the air source is turned off. The grinding chamber has a dust collection mechanism, which includes a conveying pipe and a negative pressure dust suction port. The negative pressure dust suction port is connected to the conveying pipe, which is connected to a negative pressure machine outside the grinding chamber. The grinding wheel includes a rotating shaft and a flexible sleeve, with the rotating shaft inside and the flexible sleeve outside. The sanding belt is in contact with the flexible sleeve. The tension wheel of one of the two sanding belt mechanisms faces downward. The mounting frame is equipped with a tensioning drive component, which includes a fixed part and a movable part. The tensioning wheel is located in the movable part, and the fixed part is located in the mounting frame. The mounting frame is equipped with a limit roller, the axis of which is parallel to the axis of the grinding wheel. When the limit roller touches the rail, the grinding wheel reaches the limit position of its feed into the rail. The mounting frame includes a fixed part and a movable part. The grinding wheel is located in the movable part, and the transmission wheel is located in the fixed part. When the force between the grinding wheel and the rail is greater than the force threshold, the movable part retracts. When the force between the grinding wheel and the rail is less than the force threshold, the relative positions of the movable part and the fixed part are fixed. The mounting frame is equipped with a damping mechanism, which includes a spring and a locking element. The spring is located between the movable part and the fixed part. The spring and the locking element work together to keep the movable part in a balanced position, and the distance between the grinding wheel and the transmission wheel is fixed. When grinding laterally, the axis of the grinding wheel should align with the length of the rail. During longitudinal grinding, a flap wheel is used to grind the rail web, and a first disc grinding wheel is used to grind the top surface of the rail; a second disc grinding wheel is used to grind the bottom surface of the rail; the first disc grinding wheel is used to grind the curved surface transitioning from the rail web to the rail bottom, and the second disc grinding wheel is used to grind the curved surface below the rail top that transitions to the rail web; the first disc grinding wheel forms a first angle with the horizontal plane, with the angle ranging from 0° to -10°; the second disc grinding wheel forms a second angle with the horizontal plane, with the angle ranging from 0° to 10°.

2. The rail weld grinding system as described in claim 1, characterized in that: The camera is an infrared camera and / or a visible light camera; the camera is mounted on the top of the grinding chamber.

3. The rail weld grinding system as described in claim 1, characterized in that: The sensor is a pressure sensor.

4. The rail weld grinding system as described in claim 1, characterized in that: The belt abrasive mechanism has an arc-shaped abrasive section for grinding the web of the rail and a flat abrasive section for grinding the top or bottom of the rail. The arc-shaped abrasive section is an abrasive belt tensioned on the surface of the grinding wheel, and the flat abrasive section is an abrasive belt between the grinding wheel and the tensioning wheel.

5. The rail weld grinding system as described in claim 4, characterized in that: The first sanding belt mechanism corresponds to the first trajectory, and the second sanding belt mechanism corresponds to the second trajectory. The first sanding belt mechanism and the second sanding belt mechanism are opposite each other, and the first sanding belt mechanism is higher than the second sanding belt mechanism. Based on the working position of the rail, the flat grinding part of the first sanding belt mechanism is below the arc-shaped grinding part, and the flat grinding part of the second sanding belt mechanism is above the arc-shaped grinding part.

Images