A welding quality control system based on pressure vessel manufacturing

Abstract

This invention relates to the field of intelligent industrial manufacturing technology and discloses a welding quality control system based on pressure vessel manufacturing. The system includes: setting four circumferential marks at the intersection of the nozzle and the cylinder to establish a circumferential position mapping relationship; scanning along the intersection trajectory to acquire pre-welding point clouds, establishing cross-sections at each circumferential position, extracting the nozzle normal and cylinder normal, constructing initial allowance lines, and generating a root pass trajectory; controlling the welding torch according to the normal orientation to perform controlled welding of the root pass and transition layer; scanning again when the weld cools to the interpass temperature, projecting the interpass point cloud onto the cross-section to generate actual allowance lines; calculating the remaining amount on the nozzle side and cylinder side using the actual allowance lines, controlling the welding torch to offset towards the side with the larger remaining amount based on the difference between the two sides, and performing offset welding of the fill and capping layers; finally, performing a final inspection scan and outputting a forming deviation diagram. This invention achieves compensation for cross-sectional deviations and reduces local incomplete fusion phenomena.

Description

Technical Field

[0001] This invention relates to the field of intelligent industrial manufacturing technology, and more specifically, to a welding quality control system based on pressure vessel manufacturing. Background Technology

[0002] In pressure vessel manufacturing, the saddle-shaped weld at the intersection of the shell opening and the nozzle is a critical assembly step. The weld centerline at this location is a spatial saddle-shaped curve, and the bevel opening, depth, angle between the two side walls, and heat distribution continuously change along different positions around the nozzle circumference. Currently, in multi-layer, multi-pass welding processes for such saddle-shaped bevels, conventional automated welding systems mostly plan the trajectory using a set welding torch tilt angle or a fixed theoretical centerline. During actual deposition, due to the torsional difference between the nozzle side and the shell side, and the local thermal stiffness difference on the workpiece surface, the deposited metal is easily offset towards one side wall or the bevel edge due to gravity and thermal contraction. This offset results in different cross-sectional filling requirements for the same weld at different circumferential positions. If subsequent weld passes are arranged according to a single set centerline, the local allowance deviations generated by previous weld passes will be ignored, resulting in the molten metal failing to cover the side walls and root area to be filled. As the number of weld layers increases, local deviations accumulate, easily leading to incomplete fusion defects at the root of the sidewall and the overlapping areas between layers, or causing substandard cross-sectional dimensions. Existing trajectory control methods struggle to respond promptly to changes in the actual cross-section between layers, causing the welding quality control process to lag behind the defect formation process, increasing the failure rate of subsequent non-destructive testing and rework procedures. Summary of the Invention

[0003] This invention provides a welding quality control system based on pressure vessel manufacturing, which solves the technical problems mentioned in the background art.

[0004] This invention provides a welding quality control system based on pressure vessel manufacturing, applied to welding quality control operations including a scanning head, a welding actuator with a welding torch, and a controller, for welding beveled nozzles and cylinders, comprising: The connecting pipe is clamped to the cylinder, and four circumferential marks are set outside the welding heat-affected zone to establish a mapping relationship between the circumferential position and the spatial point of the weld. Three-dimensional contour acquisition is performed along the intersecting trajectory defined by the four circumferential marks to obtain the pre-welding point cloud; Based on the pre-welding point cloud, cross sections are established at each circumferential position. The normal direction of the nozzle and the normal direction of the cylinder are extracted. The normal direction of the nozzle and the normal direction of the cylinder are used to generate the normal orientation, and the initial allowance line and the base trajectory are constructed. The welding torch of the welding actuator is controlled to perform controlled welding of the root pass and transition layer along the root pass trajectory in the normal orientation, forming a weld bead; When the weld bead cools to the interlayer temperature, the scanning head is used to collect the contour again along the intersection trajectory to obtain the interlayer point cloud, and the interlayer point cloud is projected onto the cross-section to generate the actual allowance line. Using the actual allowance line as the filling basis, extract the remaining amount on the pipe side and the remaining amount on the cylinder side within the cross-section, and control the welding torch to be offset according to the difference between the remaining amount on the pipe side and the remaining amount on the cylinder side, and perform offset welding of the remaining filling layer and the cover layer. A final inspection scan is performed on the final weld after the cover layer is completed, and the final weld contour is projected onto the cross-section to output a forming deviation diagram.

[0005] The beneficial effects of this invention are as follows: Based on the torsional change of the normal direction of the two side walls of the intersection zone, this invention compares the actual interlayer fusion result with the initial cross section to generate the actual allowance line, and dynamically adjusts the welding torch offset of the filling layer according to the difference in allowance between the pipe side and the cylinder side, overcoming the problem of local overfilling or insufficient cross-sectional size caused by welding only along the set center line; by obtaining the allowance data of the two sides through in-situ contour scanning, it realizes the directional compensation of multi-layer and multi-pass welds at different circumferential positions, suppresses the displacement of the deposited metal at the root due to gravity and thermal shrinkage, reduces the probability of the occurrence of embedded non-fusion defects, and makes the cap weld toe fall within the preset limit. Attached Figure Description

[0006] Figure 1 This is a schematic diagram of the operation scenario of the welding quality control system for the intersection area of ​​the nozzle and the cylinder of the present invention. Detailed Implementation

[0007] The subject matter described herein will now be discussed with reference to exemplary embodiments. It should be understood that these embodiments are discussed only to enable those skilled in the art to better understand and implement the subject matter described herein, and changes may be made to the function and arrangement of the elements discussed without departing from the scope of this specification. Various processes or components may be omitted, substituted, or added as needed in the examples. Furthermore, features described in some examples may be combined in other examples.

[0008] A welding quality control system based on pressure vessel manufacturing, applied to welding quality control operations including a scanning head, a welding actuator with a welding torch, and a controller, for welding beveled nozzles and cylinders, comprising: The connecting pipe is clamped to the cylinder, and four circumferential marks are set outside the welding heat-affected zone to establish a mapping relationship between the circumferential position and the spatial point of the weld. Three-dimensional contour acquisition is performed along the intersecting trajectory defined by the four circumferential marks to obtain the pre-welding point cloud; Cross-sections are established at each circumferential position according to the pre-welding point cloud, the normal directions of the nozzle and the cylinder are extracted, the normal attitude is generated from the normal directions of the nozzle and the cylinder, and an initial allowance line is constructed and a root pass trajectory is generated; Control the welding torch of the welding actuator to perform controlled welding of the root pass and the transition layer along the root pass trajectory according to the normal attitude to form a weld bead; When the weld bead cools to the interlayer temperature, use the scanning head to perform contour acquisition again along the intersection trajectory to obtain the interlayer point cloud, and project the interlayer point cloud into the cross-section to generate the actual allowance line; Taking the actual allowance line as the filling basis, extract the remaining amount on the nozzle side and the remaining amount on the cylinder side in the cross-section, control the welding torch to offset according to the difference between the remaining amount on the nozzle side and the remaining amount on the cylinder side, and perform offset welding of the remaining filling layer and the cover layer; Perform a final inspection scan on the final weld after the cover layer is completed, project the final weld contour into the cross-section, and output a forming deviation map.

[0009] S201 Fix the perforated cylinder on the roller stand or the welding positioner. The rotation accuracy of the roller stand is not less than ±0.1 degree, and the repeat positioning accuracy is not less than ±0.1 mm. Clamp both ends of the cylinder with a chuck or a pressing plate. The clamping force of the chuck is determined according to the wall thickness of the cylinder. When the wall thickness is 6 to 20 mm, the clamping force is 5 to 10 kN; when the wall thickness is 20 to 50 mm, the clamping force is 10 to 20 kN. The requirements for the arrangement of the pressing plates are as follows: for a cylinder with a diameter less than or equal to 1000 mm, the number of pressing plates is not less than 4, and the spacing is not greater than 500 mm; for a cylinder with a diameter greater than 1000 mm, the number of pressing plates is not less than 6, and the spacing is not greater than 600 mm. Adjust the rotation speed of the roller stand or the rotation angle of the welding positioner to keep the axis of the cylinder horizontal or at a preset angle of 0 to 15 degrees with the horizontal plane. The preset angle is determined according to the welding process requirements and is used to offset the influence of the gravity of the deposited metal. Use a laser level to align the levelness of the axis of the cylinder, and the levelness deviation is not greater than 0.2 mm per meter. Use a dial indicator to measure the radial runout at both ends of the cylinder, measure continuously 3 times, and each measurement result is not greater than 0.5 mm per meter as qualified. After passing the inspection, lock the braking device of the roller stand or the welding positioner, and record the fixed state of the cylinder as qualified.

[0010] The braking device adopts an electromagnetic braking method. After locking, manually rotate the cylinder, and there is no obvious displacement as the braking is qualified. The inclination angle of the cylinder is realized by adjusting the rotation angle of the welding positioner. The flat welding position adopts a 0-degree inclination, and the overhead welding position adopts an inclination of 10 to 15 degrees, which is used to offset the influence of gravity on the deposited metal.

[0011] S202. Insert the connector into the opening of the cylinder or place it on the mounting bracket on the outer surface of the cylinder. For insert-type connectors, insert the end of the connector into the opening of the cylinder to a depth of 1 to 1.5 times the connector wall thickness. The gap between the connector and the opening of the cylinder should be controlled within 0.5 to 2 mm. If the gap exceeds this range, the opening or the end of the connector needs to be adjusted. For mounting-type connectors, place the mounting bracket end face of the connector against the corresponding position on the outer surface of the cylinder, and adjust the connector axis to coincide with the center line of the opening of the cylinder. Use a dial indicator to measure the runout of the connector end face. The runout should not exceed 0.5 mm to ensure the coaxiality of the connector and the cylinder. Temporarily fix the connector by spot welding. The number of spot welds should be 4 to 6, evenly distributed around the circumference of the connector. The spot weld length should be 10 to 15 mm, and the welding current should be 100 to 120 amps. The distance between the spot weld position and the bevel area to be welded subsequently should be no less than 10 mm, and the circumferential distance from the circumferential mark should be no less than 30 mm to avoid affecting the identification of the circumferential mark. All spot weld locations should be visually inspected. The absence of cracks, porosity, and lack of fusion defects indicates a pass. Once passed, the spot weld locations should be marked with a marker. If a spot weld fails, the defects must be removed and the weld re-welded. After re-welding, the perpendicularity and coaxiality of the connecting pipe must be re-checked; the perpendicularity deviation should not exceed one-thousandth.

[0012] The perpendicularity of the connector is measured using a laser perpendicularity meter. Measurements are taken at four evenly distributed points on the connector end face, and the average value is taken as the perpendicularity deviation. The fitting gap of the saddle-type connector should not exceed 0.2 mm, and is measured using a feeler gauge. If the gap exceeds the standard, the saddle end face should be ground. The ends of the insert-type connector need to be chamfered at 30 to 45 degrees, with a chamfer depth of 1 to 2 mm, to avoid incomplete fusion during welding.

[0013] S203 uses the projection direction of the nozzle axis onto the outer surface of the cylinder as the 0-degree position, and sequentially forms 90-degree, 180-degree, and 270-degree positions along the circumference of the nozzle. A laser line projector is used to project along the nozzle axis; the intersection of the laser line with the outer surface of the cylinder and the outer surface of the nozzle is the 0-degree position reference line. Starting from the intersection of the 0-degree position reference line and the outer cylindrical surface of the nozzle, using an angle gauge or laser positioning instrument with an accuracy of no less than ±0.1 degrees, positions with angles of 90 degrees, 180 degrees, and 270 degrees from the 0-degree position are marked clockwise along the circumference of the nozzle. The four positions are evenly distributed along the circumference of the nozzle, and the angular error between adjacent positions is controlled within ±0.5 degrees. A laser rangefinder is used to measure the arc length between the 0-degree and 180-degree positions, and the arc length between the 90-degree and 270-degree positions; the difference between the two is no greater than 2 mm, verifying the symmetry of the four positions. Temporary marks are made with a marker pen, with a size no smaller than 5 mm by 5 mm.

[0014] Before use, the laser line projector must be calibrated. Place the projector on a level platform and project horizontal and vertical lines. Rotate the projector 180 degrees; a horizontal line deviation of no more than 0.2 mm per meter is acceptable. Circumferential positioning should be uniformly done clockwise; counterclockwise rotation is prohibited to ensure that the reference orientation of all workpieces is consistent.

[0015] S204 marks four circumferential marks are made on the outer surfaces of the cylinder and the nozzle at positions corresponding to 0°, 90°, 180°, and 270°. These marks are made by stamping or affixing high-temperature resistant markings. The high-temperature resistant markings must withstand temperatures no lower than 800°C and have an adhesion strength no lower than 1 MPa. Stamped marks are made using a depth-adjustable stamping tool, with a depth controlled between 0.3 and 0.5 mm and a diameter no less than 3 mm. The four circumferential marks use different shapes: circle, square, triangle, and rhombus, to facilitate automatic differentiation of marks at different positions by the scanning head. The marks are located outside the weld heat-affected zone, at a distance of no less than 50 mm from the bevel edge. One mark is made on the outer surface of the cylinder and one on the outer surface of the nozzle at each circumferential position. The line connecting the two marks is perpendicular to the nozzle axis, with a perpendicularity deviation of no more than ±1 degree. The scanning head is used to pre-scan the four circumferential marks to confirm that they are clear and identifiable, and that their shape and position meet the requirements. If a mark is not identifiable, the original mark is removed and a new one is made.

[0016] Before affixing the high-temperature resistant markings, wipe the surface with acetone to remove oil and dust. After affixing, press firmly with your finger for 30 seconds to ensure a secure bond. After the stamped markings are completed, use fine sandpaper to smooth out any burrs around the markings, covering an area of ​​about 2 millimeters to avoid stress concentration.

[0017] The S205 uses a controller to read the spatial coordinates of four circumferential marks and establish a mapping relationship. During the system debugging phase, hand-eye calibration was completed using a nine-point calibration method. The calibration plate was placed within the welding workspace, and the scanning head and welding torch respectively acquired nine feature points from the calibration plate, calculating the homogeneous transformation matrix between them. The calibration accuracy was no greater than 0.1 mm. The controller was connected to a 3D scanning device with an accuracy of no less than ±0.1 mm, and the spatial coordinates of the four circumferential marks in the device coordinate system were acquired sequentially. A workpiece coordinate system was established, with the origin at the projection point of the circumferential mark on the outer surface of the cylinder at the 0-degree position. The X-axis was along the cylinder axis, the Y-axis along the nozzle axis, and the Z-axis perpendicular to the plane formed by the cylinder axis and the nozzle axis. The transformation matrix obtained from the hand-eye calibration was used to convert the mark coordinates in the scanning head coordinate system to coordinates in the workpiece coordinate system. Using the coordinates of four circumferential markers as a reference, a one-to-one correspondence is established between circumferential angles and spatial points of the weld. The intersection trajectory of the weld is discretized according to the circumferential angles, with a discretization step size not exceeding 1 degree. Each discrete point corresponds to a circumferential angle value, forming a mapping relationship table. Outside of the 0°, 90°, 180°, and 270° positions, three intermediate positions are randomly selected, and their actual spatial coordinates are measured against the coordinates calculated from the mapping relationship. The difference should not exceed 0.2 mm to verify the accuracy of the mapping relationship. If the mapping relationship fails to be established, such as due to excessive error in the marker coordinate acquisition, the coordinates of the four circumferential markers are reread, and the mapping relationship is re-established. Pre-welding scanning must be initiated within 4 hours of establishing the mapping relationship. If more than 4 hours have passed, the coordinates of the four circumferential markers must be reread, and the reference deviation must be verified to be no greater than 0.1 mm before continuing. If the workpiece displacement exceeds 0.5 mm, the welding process is interrupted for more than 2 hours, or the equipment is restarted, the mapping relationship must be re-established.

[0018] The mapping table includes four fields: circumferential angle, X, Y, and Z coordinates of the corresponding weld space point, and weld tangential vector. Nine-point calibration uses a 100mm x 100mm ceramic calibration plate, with the nine feature points evenly distributed in a 3x3 pattern and a 30mm spacing between adjacent points. During calibration, the scanning head and welding torch separately acquire the center coordinates of each feature point, and the average value is calculated after three acquisitions.

[0019] S301 The bevel at the intersection of the connecting pipe and the cylinder is cleaned. Use a stainless steel wire brush or an angle grinder wheel with an 80-120 grit to polish the bevel surface and both sides, removing scale, oil, rust, and other impurities. The wire diameter of the wire brush should not exceed 0.3 mm. The polishing area includes the inner and outer walls of the bevel, as well as an area of ​​at least 30 mm on both sides of the bevel edge. For stainless steel pressure vessels, after cleaning, wipe the bevel surface with acetone to remove residual grease and carbon contamination; for carbon steel pressure vessels, welding must be performed within 24 hours after cleaning to prevent re-rusting. After cleaning, blow the bevel surface with clean compressed air to remove residual dust and debris. Wipe the bevel surface with a white silk cloth; the cloth should be free of obvious stains to indicate acceptance. After acceptance, cover the bevel area with a dust cover to prevent dust and debris contamination.

[0020] The compressed air pressure is controlled between 0.4 and 0.6 MPa, the purging distance is 100 to 200 mm, and the purging angle is 45 to 60 degrees, purging from the root of the bevel outwards. The dust cover is made of non-woven fabric with a thickness of 0.1 mm, covering the bevel and 50 mm areas on both sides, and is fixed to the cylinder and connecting pipe surfaces with tape.

[0021] The S302 moves the scanning head to the circumferential mark corresponding to the 0-degree position and reads the relative position of the circumferential mark and the bevel edge. The controller calculates the spatial coordinates of the circumferential mark at the 0-degree position based on the mapping table and automatically controls the scanning head to move to a preset height of 150 to 250 mm above this coordinate. The scanning head's posture is adjusted so that the angle between the scanning head's axis and the weld tangent is 90 degrees ± 1 degree, and the working surface of the scanning head is parallel to the weld surface. The scanning angle of the scanning head is ± 30 degrees to ensure complete coverage of the bevel area. When the ambient temperature changes by more than ± 5 degrees Celsius, the scanning head needs to be calibrated for temperature by scanning a standard gauge block to correct systematic errors in the scanning distance. The scanning head acquires point cloud data for this area, and the controller extracts the center coordinates of the circumferential mark and the coordinates of both sides of the bevel from the point cloud data. The center coordinates of the circumferential mark are the centroid coordinates of all point clouds in the marked area, and the bevel edge coordinates are the coordinates of the points with the maximum gradient change in the point clouds on both sides of the bevel. The relative distance and relative angle between the circumferential mark and the bevel edge are calculated. Compare the read relative position with the theoretical value; a deviation of no more than 0.2 mm is considered acceptable. Only after passing the test can 360-degree continuous scanning be started. If the reading fails, adjust the position and orientation of the scanning head and try reading again.

[0022] During scanning head temperature calibration, the standard gauge block is scanned at three positions: left, center, and right. 100 points are collected at each position, the average scanning distance is calculated, and compared with the actual size of the gauge block to obtain the systematic error. Then, all scanning distances are corrected. The point of maximum gradient change is obtained by calculating the coordinate difference between adjacent points; the point with the largest difference is the bevel edge point.

[0023] The S303 controller moves the scanning head sequentially along the circumference of the connector, passing through 90-degree, 180-degree, and 270-degree positions before returning to the 0-degree position, completing 360-degree continuous data acquisition. The controller sends a scan start command to the scanning head, which then begins continuous acquisition along the intersecting trajectory at a preset speed of 50 to 100 mm per second. During scanning, the distance between the scanning head and the weld surface is kept constant, with an allowable deviation of ±10 mm. The minimum safe distance between the scanning head and the workpiece is 50 mm; if the scanning head detects a distance less than 50 mm, it immediately stops and issues an alarm signal. Upon passing each circumferential marker, an iterative nearest-point algorithm is used to register the currently scanned marker point cloud with the initial marker point cloud, calculating the scanning head's positional deviation, and then real-time correcting the subsequent scanning trajectory. If scanning is interrupted, the scanning head is moved to the nearest circumferential marker, recalibrated, and then scanning resumes from the interrupted position, ensuring the continuity of the point cloud data. The scanning head passes through the circumferential marks at 90-degree, 180-degree, and 270-degree positions in sequence, and finally returns to the 0-degree position, completing a 360-degree continuous scan of the entire intersection trajectory.

[0024] The registration threshold for the iterative closest point algorithm is 0.05 mm. Iteration stops when the registration error is less than the threshold. The alarm signal is in the form of an audible and visual alarm: a red indicator light flashes, and a buzzer emits intermittent sounds. The alarm information includes the fault location as abnormal scanning head movement and the fault cause as being too close to the target.

[0025] The S304 scanner acquires cross-sectional profiles of the outer wall of the nozzle, the root of the bevel, and the outer wall of the cylinder at various circumferential positions. As the scanning head moves along the intersection trajectory, the controller calculates the current circumferential angle in real time based on a mapping table. A cross-sectional acquisition is triggered every 1-degree preset acquisition interval. For U-shaped bevels, the acquisition interval can be increased to 1.5 degrees; for V-shaped bevels, the interval remains no greater than 1 degree. The scanning range of each cross-section covers the outer wall of the nozzle, the entire bevel area, and the outer wall of the cylinder, ensuring that the cross-sectional profile completely encompasses all boundaries of the area to be welded. At least 100 points are acquired for each cross-section, including at least 20 points on the outer wall of the nozzle, 30 points in the bevel area, and 20 points on the outer wall of the cylinder; otherwise, the cross-section is re-acquired. Reflective points exceeding a threshold intensity are directly deleted, and then interpolated using corresponding points from adjacent cross-sections to complete the profile, preventing reflective points from affecting the accuracy of the profile extraction. The acquired cross-sectional point cloud data is stored in circumferential angle order, with each cross-sectional data point corresponding to a unique circumferential angle value.

[0026] The intensity threshold for reflective points is set to 1.5 times the average intensity of normal point clouds; points with intensity exceeding the threshold are identified as reflective points. The bevel shape is identified by the bevel angle and root radius; a root radius greater than 2 mm is classified as a U-shaped bevel, and a root radius less than or equal to 2 mm is classified as a V-shaped bevel.

[0027] The S305 uses the acquired point cloud as the pre-welding point cloud. The controller stitches together the point cloud data from all cross-sections in ascending order of circumferential angles, ensuring an overlap of at least 10% between adjacent cross-sections. An iterative nearest-point algorithm is used for point cloud stitching, resulting in an overall point cloud error of no more than 0.1 mm. A statistical filtering algorithm is employed to remove noise and outliers from the point cloud data. The filtering parameters are automatically adjusted based on the point cloud density: when the density is greater than 100 points per square centimeter, the number of neighboring points is 20, and the standard deviation factor is 2; when the density is less than or equal to 100 points per square centimeter, the number of neighboring points is 15, and the standard deviation factor is 1.5. After stitching and denoising, a complete pre-welding point cloud is obtained. The pre-welding point cloud is stored in polygon or point cloud data format, with coordinate precision retained to three decimal places. The file name is the workpiece number followed by "pre-welding," then the year, month, day, hour, minute, second, and the corresponding format suffix, and stored in the pre-welding point cloud folder specified by the controller. The integrity and density of the pre-welding point cloud are then checked. If the point cloud quality is not up to standard, perform a pre-welding scan again.

[0028] The overlapping area of ​​the point cloud stitching is taken as the average value of the point clouds of two adjacent sections to eliminate stitching errors. The pre-welding point cloud is stored in the "Pre-welding Point Cloud" folder under the "Welding Data" drive on the controller. Subfolders are created according to the workpiece number, and all data for each workpiece is stored in the corresponding subfolder.

[0029] S401 fits the outer cylindrical surfaces of the cylinder and the nozzle from the pre-welding point cloud, extracting the cylinder normal and nozzle normal at corresponding circumferential positions. Point cloud data for the outer surfaces of the cylinder and nozzle are filtered based on their geometric features. The filtering criteria for the cylinder outer surface point cloud are: axial coordinates within the cylinder length range, and radial coordinates within ±5 mm of the cylinder radius; the filtering criteria for the nozzle outer surface point cloud are: axial coordinates within the nozzle length range, and radial coordinates within ±5 mm of the nozzle radius. The least squares method is used to fit the filtered cylinder outer surface point cloud to a cylindrical surface, obtaining the cylinder's cylindrical surface equation; similarly, the least squares method is used to fit the filtered nozzle outer surface point cloud to a cylindrical surface, obtaining the nozzle's cylindrical surface equation. The mean square error of the cylindrical surface fitting is no greater than 0.1 mm, and the number of iterations does not exceed 100. If the error exceeds the threshold, the point cloud is re-filtered and fitted. For each circumferential position, at the intersection of the cross-section at that position and the outer cylindrical surface of the cylinder, the normal vector of the cylinder's cylindrical surface at that point is calculated; this is the cylinder normal. Similarly, at the intersection of the cross-section at that position and the outer cylindrical surface of the nozzle, the normal vector of the nozzle's cylindrical surface at that point is calculated; this is the nozzle normal. Thirty neighborhood points with a radius of 5 mm are used for calculating the normals. Using the cylinder center and nozzle center as references, all normal vectors are adjusted to point outwards from the base material. Normal vectors with opposite directions are multiplied by -1 for correction. The cylinder normal and nozzle normal data for each circumferential position are bound to the circumferential angle at that position and stored accordingly.

[0030] When fitting a cylindrical surface using the least squares method, an objective function is established to sum the squares of the distances from a point to the cylindrical surface. The radius and center coordinates of the cylindrical surface are then solved iteratively. To normalize the normal vector, each component of the vector is divided by its magnitude to obtain the unit normal vector.

[0031] S402 extracts the weld tangent along the weld center region, establishing a cross-section perpendicular to the weld tangent. The weld center region is a 10 mm area extending inward from both sides of the bevel edge; the point cloud within this region is used to extract the weld centerline. The weld centerline is extracted using the bevel edge midpoint method; that is, the line connecting the midpoints of the two bevel edges within each cross-section is the weld centerline. The centerline point cloud is smoothed using a moving average with a window size of 5 points. Symmetrical boundary conditions are used at the start and end positions of the weld for smoothing to avoid abrupt changes in trajectory at the boundaries. The tangent direction of the centerline at each circumferential position is calculated, which is the weld tangent. With the weld center at that circumferential position as the origin and the weld tangent as a coordinate axis, a plane perpendicular to the weld tangent is established; this plane is the cross-section at that circumferential position. Each cross-section is a thin region with a thickness of 0.1 mm, and all point clouds within this region are projected onto the cross-sectional plane. All cross-sections are perpendicular to the weld tangent at the corresponding position and are uniformly distributed along the weld centerline.

[0032] The weld tangential direction is obtained by dividing the coordinate difference between two adjacent points on the centerline by the distance between the two points. The cross-sectional thickness is determined based on the point cloud density: 0.1 mm when the point cloud density is greater than 100 points per square centimeter, and 0.2 mm when the point cloud density is less than or equal to 100 points per square centimeter.

[0033] S403 defines the bevel's sidewall line (closer to the nozzle) and the sidewall line (closer to the cylinder) within the cross-section as the nozzle sidewall line, and also defines the root contour line. All point clouds within a 0.1 mm thickness range around this cross-section are projected onto the cross-sectional plane to obtain a two-dimensional point cloud set. If the cross-section and point cloud do not intersect, the cross-section position is adjusted, and the projection is repeated. Edge detection is performed on the two-dimensional point cloud to extract the bevel's boundary contour. The edge line of the bevel near the nozzle is defined as the nozzle sidewall line, using linear fitting with a mean square error (MSE) of no more than 0.05 mm; the edge line of the bevel near the cylinder is defined as the cylinder sidewall line, using linear fitting with an MSE of no more than 0.05 mm; the contour line at the bottom of the bevel is defined as the root contour line, using quadratic curve fitting with an MSE of no more than 0.05 mm. The lowest point of the root contour line is the vertex of the fitted quadratic curve, representing the deepest point of the bevel. When the bevel edge point cloud is incomplete, the fitted straight line or curve is extended outward by 1 mm to ensure the continuity of the boundary line. Check the continuity and closure of the three boundary lines. A closed bevel region formed by the three lines is considered acceptable. If the boundary lines are discontinuous, use interpolation to complete them before proceeding with further processing. The coordinate error of the extracted boundary lines should not exceed 0.1 mm.

[0034] Edge detection employs the Sobel operator to calculate gradients in both the horizontal and vertical directions, synthesizing the gradient magnitude. The point with the largest magnitude is identified as the edge point. The root contour is fitted using a second-order polynomial, with the fitting interval being a 5 mm range on each side of the bevel root.

[0035] S404 uses the nozzle sidewall line and the cylinder sidewall line as boundaries to decompose the area to be filled within the cross-section into initial nozzle sidewall allowance, initial root allowance, and initial cylinder sidewall allowance. Within the cross-section, the area to be filled is a closed area enclosed by the nozzle sidewall line, the cylinder sidewall line, and the root contour line. An auxiliary line is drawn from the midpoint of the root contour line towards the bevel opening direction. The direction of the auxiliary line is the angle bisector of the cylinder normal and the nozzle normal, and the length of the auxiliary line is equal to the bevel depth. Using the auxiliary line as a boundary, the area to be filled is divided into three parts. The part between the nozzle sidewall line and the auxiliary line is defined as the initial nozzle sidewall allowance; the part between the auxiliary line and the cylinder sidewall line is defined as the initial cylinder sidewall allowance; and the part between the root contour line and the bottom of the auxiliary line is defined as the initial root allowance. The initial root allowance area is the area within 2 mm on each side of the root contour line and the bottom of the auxiliary line; this area is the filling range for the root pass weld. The polygon area calculation method is used to connect the boundary points of the area to be filled in sequence to form a polygon, and then calculate the area of ​​the polygon to obtain the area value of each initial allowance. When the bevel angles of the pipe side and the cylinder side are different, the auxiliary line is still along the direction of the double normal angle bisector to ensure that the allowances on both sides after decomposition are consistent with the actual filling requirements.

[0036] The area of ​​the polygon is calculated using the shoelace method. The boundary points are arranged clockwise, and the sum of the coordinate products of adjacent points is calculated. Half of the absolute value is the area. The auxiliary lines are connected to the midpoints of the root contour line with a straight line, and the connection point is the lowest point of the root contour line.

[0037] S405 connects the initial nozzle-side allowance, initial root allowance, and initial cylinder-side allowance in circumferential order to form the initial allowance line. For each circumferential position cross-section, the upper edge points of the initial nozzle-side allowance, initial root allowance, and initial cylinder-side allowance are extracted respectively. The upper edge point is the intersection of each allowance region with the bevel opening edge; that is, the upper edge point of the initial nozzle-side allowance is the intersection of the nozzle sidewall line and the bevel opening edge, the upper edge point of the initial cylinder-side allowance is the intersection of the cylinder sidewall line and the bevel opening edge, and the upper edge point of the initial root allowance is the intersection of the auxiliary line and the bevel opening edge. The coordinate error of the upper edge point is no greater than 0.05 mm. Connect the above three edge points at all circumferential positions in ascending order of circumferential angle, and smooth them using cubic spline interpolation to form three continuous curves. Merge these three curves to obtain the initial allowance line distributed along the entire intersection trajectory. The discrete point interval of the initial allowance line should be consistent with the cross-section acquisition interval, i.e., no greater than 1 degree. The initial allowance line should be compared with the theoretical bevel profile in the computer-aided design model; a deviation of no more than 0.3 mm is acceptable. If the deviation exceeds 0.3 mm, pre-welding scanning and point cloud processing should be repeated.

[0038] Cubic spline interpolation uses natural boundary conditions, with the second derivative at both endpoints being zero. The theoretical bevel profile is derived from the computer-aided design model in a two-dimensional curve coordinate point set with a point interval of 1 degree.

[0039] The S406 extracts the angle bisector between the nozzle normal and the cylinder normal as the normal attitude. Within the cross-section at each circumferential position, the vectors of the cylinder normal and nozzle normal are normalized. The two unit vectors are then added, and the sum vector is normalized again. The resulting unit vector is the direction of the angle bisector between the two normals. This angle bisector direction represents the normal attitude of the welding torch at that circumferential position, and the torch axis points inwards along this direction towards the bevel. When the angle between the cylinder normal and the nozzle normal is 180 degrees, the angle bisector direction is perpendicular to the plane formed by the cylinder axis and the nozzle axis. Using an inverse kinematics algorithm, the normal attitude vector is converted into the six joint angles of the welding robot, with a conversion error of no more than 0.1 degrees. During system debugging, the relative positional relationship between the scanning head and the welding torch is obtained through hand-eye calibration, and then the normal attitude is compensated, with a compensation accuracy of no more than 0.1 mm. The deviation of the normal attitude from the theoretical angle bisector is no more than ±1 degree. The normal orientation of each circumferential position is bound and stored with the discrete point of the initial margin line at that position.

[0040] The inverse kinematics of the robot is solved using the DH parameter method. A coordinate system is established for the six links of the robot, and the angles of each joint are calculated based on the pose of the end effector. During installation error compensation, the transformation matrix obtained from hand-eye calibration is multiplied by the normal attitude vector to obtain the compensated welding torch attitude vector.

[0041] The S407 generates a root pass trajectory based on the geometric center and normal orientation of the initial root allowance. Within the cross-section of each circumferential position, the geometric center coordinates of the initial root allowance region are calculated using the centroid method, which is the average coordinate of all pixels in the initial root allowance region. The height of the root pass trajectory from the root of the bevel is 1 to 2 mm to ensure complete penetration of the root pass. The geometric center coordinates of all circumferential positions are connected in ascending order of circumferential angle to form a spatial curve, which is the centerline of the root pass trajectory. The discrete point interval of the root pass trajectory is no greater than 0.5 mm to ensure the smoothness of the welding torch movement. Cubic spline interpolation is used to smooth the root pass trajectory, ensuring the continuity of the first and second derivatives. The orientation of the root pass trajectory is determined by the normal orientation at the corresponding position, ensuring that the axis of the welding torch always points towards the root of the bevel along the normal orientation when moving along the root pass trajectory. The root pass trajectory is converted into a motion command format recognizable by the welding robot and sent to the robot controller. The robot controller first performs trajectory simulation to verify that the trajectory is collision-free and the posture is continuous. Only if the simulation is successful can the robot be executed. If the simulation fails, a new baseline trajectory is generated.

[0042] When calculating the geometric center using the centroid method, the X coordinates of all points within the initial root margin region are summed and divided by the number of points to obtain the X coordinate. Similarly, the Y and Z coordinates are obtained. During trajectory simulation, the minimum distance between the welding torch and the workpiece and fixture is checked. If the minimum distance is less than 10 mm, it is considered a collision, and the trajectory height is adjusted until the minimum distance is greater than 10 mm.

[0043] S501 moves the welding torch to the preset arc-starting point, aligning the torch's axis with the normal orientation of the corresponding circumferential position. The preset arc-starting point is set at the weld's starting position, 10 to 15 millimeters above the root of the bevel at the 0-degree position. The lateral position of the arc-starting point is the geometric center of the initial root allowance area, ensuring accurate arc initiation. The controller moves the welding actuator to the preset arc-starting point, adjusting the torch's orientation so that its axis coincides with the normal orientation of that circumferential position. The distance between the torch nozzle and the root of the bevel is maintained at 8 to 12 millimeters to ensure stable arc combustion. A laser tracker measures the actual position and orientation of the torch nozzle. Comparing this to the theoretical values, a positional deviation of no more than 0.2 millimeters and an orientation deviation of no more than 0.1 degrees are considered acceptable. Welding power can only be started after this is deemed acceptable. If the positioning is unacceptable, the torch's position and orientation must be readjusted.

[0044] During laser tracking measurements, a 12.7 mm diameter ceramic target ball is installed at the end of the welding torch nozzle to collect the spatial coordinates of the target ball. Ten consecutive measurements are taken, and the average value is calculated. The welding torch is aligned using a trial welding method. A 100 mm long weld is welded onto a test plate made of the same material as the workpiece. The plate is then cut perpendicular to the weld seam, ground, and polished. The deviation between the weld seam center and the bevel center is measured; a deviation of no more than 0.2 mm is considered acceptable.

[0045] The S502 starts the arc and advances along the intersecting trajectory, reads the corresponding initial allowance line, and restricts the lateral position of the welding torch within the initial root allowance to perform root pass fusion. The controller sends a start command to the welding power source, ignites the arc, and then controls the welding torch to advance at a constant speed of 100 to 150 mm per minute along the intersecting trajectory. The welding parameters are 100 to 150 A current and 20 to 25 V voltage. During the advance, the controller reads the initial allowance line data corresponding to the current circumferential position in real time and adjusts the lateral oscillation range of the welding torch according to the width of the initial root allowance. The welding torch uses a linear oscillation mode, with the oscillation direction perpendicular to the welding direction, the oscillation center coinciding with the root pass trajectory, the oscillation width being 80% to 90% of the initial root allowance width, the oscillation frequency being 0.5 to 1 Hz, and the dwell time on both sides being 0.2 to 0.5 seconds to ensure good sidewall fusion. The penetration state is controlled by the feedback signal of the welding current. When the current exceeds the preset threshold by 10%, the welding speed is increased appropriately; when the current is lower than the preset threshold by 10%, the welding speed is decreased appropriately. The height of the root pass after completion is 2 to 3 mm, which is 1 to 2 mm lower than the root of the bevel, leaving space for subsequent transition layer welding.

[0046] The welding current feedback signal is sampled 100 times per second, and noise is removed using a moving average filter. The height of the root pass weld is measured using a depth gauge, with 8 measurement points evenly selected around the weld circumference, and the average value is taken as the weld height.

[0047] As the S503 passes four circumferential marks, it synchronizes the welding torch's posture with the normal directions of the cylinder and nozzle within the cross-section. During welding, the controller calculates the welding torch's current circumferential position in real time. When the distance to a circumferential mark is less than 0.5 degrees, it begins to gradually adjust the welding torch's posture, completing the posture switch upon reaching the mark. The welding torch's rotational speed is proportional to the welding speed, with a proportionality coefficient equal to the posture change divided by the welding distance, ensuring synchronization between posture changes and the welding process. The acceleration of posture changes does not exceed 50 degrees per second squared to avoid impact from the welding torch's movement, which could affect welding stability. When the difference in normal posture between two adjacent circumferential marks exceeds 10 degrees, the posture change is divided into five transition segments, each with a change of no more than 2 degrees. Between two adjacent circumferential marks, the welding torch's posture transitions smoothly using linear interpolation, with an interpolation step no greater than 0.1 degrees. The posture rotation is performed in the workpiece coordinate system to ensure that the welding torch's posture remains correct relative to the workpiece.

[0048] The attitude change is the angle between the normal attitude vectors of two adjacent circumferential positions, calculated through the vector dot product. For linear interpolation, the joint angles at the two marked positions are interpolated based on the ratio of the current circumferential angle to the angles at the two marked positions to obtain the joint angle at the current position.

[0049] After the initial weld penetration and arc termination, the S504 performs controlled welding of the transition layer within the same coordinate system. The centerline of the transition layer is obtained by translating the initial allowance line towards the bevel edge, ensuring that the endpoints of the transition layer do not exceed the sidewall lines of the nozzle and cylinder. After the initial weld penetration and arc termination, the welding torch remains at the termination point for 2 to 3 seconds to fill the crater. Then, the controller automatically switches the welding parameters to the transition layer welding parameters: current 120 to 180 amps, voltage 22 to 28 volts, and welding speed 120 to 180 mm per minute. The number of transition layers is determined by the bevel depth. When the bevel depth is greater than 5 mm, two transition layers are welded; when the bevel depth is less than or equal to 5 mm, one transition layer is welded. The centerline of the transition layer is obtained by translating the initial allowance line towards the sidewall lines of the nozzle and cylinder, respectively, by 4 to 6 mm. The translation distance is half the width of the transition layer weld bead, and the width of the transition layer weld bead is 8 to 12 mm. The overlap width between the transition layer weld and the root pass weld should be 30% to 40% of the root pass weld width to ensure good interlayer fusion. When welding the transition layer, control the range of motion of the welding torch so that the two ends of the transition layer are located inside the pipe sidewall line and the cylinder sidewall line, respectively, and do not exceed the positions of the sidewall lines. After the transition layer welding is completed, the welding torch should automatically return to a safe position. Visually inspect the surface of the transition layer; it is acceptable if there are no cracks, porosity, slag inclusions, or other defects, and the weld height is uniform with a height difference not exceeding 0.5 mm. If defects are present, remove the defects and re-weld the transition layer.

[0050] The crater filling method employs current decay, where the welding current linearly decreases to 50 amps within 0.5 seconds upon arc termination, and is then held for 0.3 seconds before extinguishing the arc. The width of the transition layer weld bead is measured using vernier calipers, with eight measurement points evenly selected around the weld circumference, and the average value taken as the weld bead width.

[0051] After the controlled welding of the root pass and transition layer is completed and the weld has cooled to the interpass temperature, S601 performs interpass cleaning on the weld surface. The controller starts the cooling timer program, and a contact thermometer is used to monitor the temperature of the weld surface in real time. The measurement position is the highest temperature point of the weld, i.e., 100 mm behind in the welding direction. For commonly used pressure vessel steel Q345R, the interpass temperature is controlled between 100 and 200 degrees Celsius. An alert signal is issued after the weld temperature has cooled to the interpass temperature range before interpass cleaning can begin. A stainless steel wire brush is used to grind the weld surface to remove slag, spatter, and oxide scale. Carbon steel wire brushes are prohibited from being used to clean stainless steel welds to avoid carbon contamination. Abrasive wheels are prohibited from being used to grind the transition layer surface unless there are obvious weld beads or defects. The grinding area includes the entire weld surface and the bevel edges on both sides. After cleaning, the weld surface is blown clean with compressed air to remove residual debris. The cleaned surface is wiped with a white silk cloth; the absence of obvious stains on the cloth indicates acceptance.

[0052] The measurement accuracy of the contact thermometer is plus or minus 1 degree Celsius. When measuring, press the temperature measurement probe vertically on the surface of the weld bead with a pressing force of 0.5 to 1 N, and read the temperature value after holding for 2 seconds. After interlayer cleaning, check the surface of the weld bead with a 10x magnifying glass. It is qualified if there is no residual welding slag and scale.

[0053] S602 Move the scanning head to the 0-degree position and read the spatial coordinates of the four circumferential marks again. The controller calculates the spatial coordinates of the circumferential marks at the 0-degree position according to the mapping relation table and automatically controls the scanning head to move to this position. Re-collect the spatial coordinates of the four circumferential marks, and the reading accuracy is not greater than 0.1 mm. Compare the coordinates collected this time with the initial coordinates collected before welding and calculate the displacement deviation. Decompose the displacement deviation into translational deviations in the X, Y, and Z directions and rotational deviations around the X, Y, and Z axes. Use the rigid body transformation algorithm to correct the subsequent scanning and welding trajectories. Measure the coordinates of the marks at the 0-degree position, and the deviation from the initial coordinates is not greater than 0.1 mm to verify the accuracy of the corrected coordinate system. If the displacement deviation exceeds 0.5 mm, re-establish the mapping relation. After the correction is completed, the interlayer scanning can be started.

[0054] The rigid body transformation algorithm obtains the translation vector and rotation matrix by solving the least squares problem. After the coordinate system is corrected, if the verification is unqualified, re-read the coordinates of the four circumferential marks and perform correction again. If the correction is unqualified three times consecutively, re-establish the mapping relation.

[0055] S603 Collect the contours of the surface of the weld bead, the remaining area of the groove on the nozzle side, and the remaining area of the groove on the shell side along the intersection trajectory as the interlayer point cloud. The parameters such as the movement speed, scanning distance, and cross-section acquisition interval of the interlayer scanning are exactly the same as those of the pre-welding scanning to ensure the comparability of the point cloud data. The range of the interlayer scanning extends 5 mm to both sides compared with the pre-welding scanning to ensure the complete acquisition of the contours of the weld bead surface and the remaining groove area. The scanning head performs a 360-degree continuous scan along the intersection trajectory, and position calibration is performed at each circumferential mark during the scanning process. The interlayer point cloud is stitched based on the re-read circumferential mark coordinates, using the same coordinate system as the pre-welding point cloud, and the registration accuracy is not greater than 0.1 mm. The denoising parameters of the interlayer point cloud are the same as those of the pre-welding point cloud, but for the convex points on the weld bead surface, their original data is retained and no denoising treatment is performed. If the point cloud is missing at a certain position, re-collect the cross-section data at this position.

[0056] The registration of the interlayer point cloud and the pre-welding point cloud adopts the iterative closest point algorithm, and the registration threshold is 0.1 mm. The method for determining the convex point is that if the height of a certain point is more than 1 mm higher than the average height of the surrounding 10 points, it is determined as a convex point and the original data is retained.

[0057] S604 projects the interlayer point cloud onto the cross-section to extract the actual weld line at each circumferential position. The interlayer point cloud is then projected onto each pre-existing cross-section to obtain the actual two-dimensional contour of the weld bead within each cross-section. A threshold-based segmentation method is used to extract the weld bead surface point cloud. Since the intensity of the weld bead surface point cloud is higher than that of the bevel surface point cloud, a strength threshold is set to filter out the bevel surface point cloud and the slag point cloud. The extracted weld bead surface point cloud is fitted to obtain the upper edge line of the weld bead, which is the actual weld line at that circumferential position. The actual weld line is processed using a moving average smoothing algorithm with a window size of 3 points. The extraction error of the actual weld line is no greater than 0.1 mm. Cubic spline interpolation is used between two adjacent cross-sections to obtain a continuous actual weld line, with an interpolation error no greater than 0.05 mm.

[0058] The threshold for threshold segmentation is the average strength of 100 random points on the weld surface. The cubic spline interpolation error is the difference between the coordinates of the interpolated point and the actual point; an error of no more than 0.05 mm is considered acceptable.

[0059] S605 generates the actual allowance line by replacing the original contour at the corresponding position in the initial allowance line with the actual weld line. For each circumferential position of the cross section, the actual weld line at that position is aligned with the initial allowance line at that position, with an alignment error not exceeding 0.1 mm. The original contour portion at the corresponding position in the initial allowance line is replaced with the actual weld line, covering an area half a cross section interval before and after that circumferential position. All contours after replacement at all circumferential positions are spliced ​​using cubic spline interpolation to form a continuous actual allowance line. The continuity and rationality of the actual allowance line are checked; no abrupt changes or intersections are considered acceptable. The first derivative of the actual allowance line is continuous, and the rate of change of the second derivative is not greater than 0.1. When the actual weld line exceeds the initial allowance line, the remaining amount is recalculated using the actual weld line as the new boundary, and an overfill warning signal is issued. The actual allowance line must be updated after each layer is welded.

[0060] Overfill warning signals are displayed as a red pop-up on the controller screen, showing the circumferential location and amount of overfill, and are also recorded in the welding log. After the actual allowance line is updated, three circumferential locations are randomly selected, and the deviation between the actual weld line and the initial allowance line is compared. A deviation of no more than 0.3 mm is considered acceptable.

[0061] S701 calculates the remaining amount on the nozzle side and the remaining amount on the cylinder side within the cross-section at each circumferential position based on the actual allowance line. Within the cross-section at each circumferential position, the remaining area to be filled is defined by the actual allowance line as the lower boundary, the nozzle sidewall line and the cylinder sidewall line as the left and right boundaries, and the designed upper edge of the bevel as the upper boundary. The designed upper edge of the bevel is the line connecting the upper endpoints of the nozzle sidewall line and the cylinder sidewall line. The area on the nozzle side of the remaining area to be filled is calculated; the area on the cylinder side is calculated. The unit of the remaining amount is square millimeters, and the calculation accuracy is retained to two decimal places. After each layer is welded, the actual allowance line is updated with the actual weld line of that layer, and then the remaining amount is recalculated. The remaining amount is the area of ​​the area to be filled above the current actual allowance line. When the remaining amount is negative, it indicates that the position has been overfilled. During subsequent welding, the welding torch is offset to the other side to reduce the amount of deposited metal on that side.

[0062] The upper edge of the bevel design is derived from the computer-aided design model or calculated based on the dimensions of the bevel machining drawings. When the remaining amount is negative, the offset distance is proportional to the absolute value of the remaining amount, and the proportionality coefficient is the same as when it is positive. The maximum offset distance does not exceed 3 mm.

[0063] For each remaining fill layer, S702 determines the relationship between the remaining amount on the nozzle side and the remaining amount on the cylinder side. It generates a welding trajectory close to the corresponding sidewall line on the side with the larger value, and then generates an overlap trajectory towards the center of the cross-section. The remaining amount data is sent to the offset trajectory generation module. The module first determines whether the current remaining amount is less than or equal to the fill amount required for the capping layer (the area corresponding to a height of 2 to 3 mm). If so, it generates a capping layer welding trajectory; otherwise, it generates the welding trajectory for the next fill layer. For each remaining fill layer, it iterates through the remaining amounts on the nozzle side and cylinder side at all circumferential positions. For each circumferential position, it compares the values ​​of the two remaining amounts. If the remaining amount on the nozzle side is greater than the remaining amount on the cylinder side, it first generates a welding trajectory close to the nozzle sidewall line, and then generates an overlap trajectory from this trajectory towards the center of the cross-section; if the remaining amount on the cylinder side is greater than the remaining amount on the nozzle side, it first generates a welding trajectory close to the cylinder sidewall line, and then generates an overlap trajectory towards the center of the cross-section. The overlap width between the lap track and the previous weld bead is 30% to 40% of the weld width. The height of each weld bead is 2 to 3 millimeters. The number of weld bead tracks is equal to the difference between the remaining width of the area to be filled and the weld bead width minus the overlap width, and the result is rounded up. The lap track is offset from the weld bead track close to the sidewall towards the center of the cross-section by a distance equal to the weld bead width minus the overlap width.

[0064] The required filling amount for the cover layer is the cover height multiplied by the groove width, and the cover height is taken as 2.5 mm. Example for calculating the number of deposition tracks: with a groove width of 20 mm, a bead width of 8 mm, and an overlap width of 3 mm, the number of tracks is 20 divided by 5, which equals 4, and rounded up to 4 tracks.

[0065] S703 performs offset welding for the remaining filling layers in the circumferential order. When passing through the position where the remaining amount on the nozzle side is greater than that on the shell side, the welding torch is controlled to be offset towards the side wall line of the nozzle; when passing through the position where the remaining amount on the shell side is greater than that on the nozzle side, the welding torch is controlled to be offset towards the side wall line of the shell. The welding parameters for the filling layer are a current of 150 to 220 A, a voltage of 24 to 32 V, and a welding speed of 150 to 200 mm per minute. The robot controller simulates the generated deposition tracks to verify that there are no collisions and the offset amount is continuous. After passing the simulation, the corresponding welding parameters are loaded to start the offset welding. The offset distance of the welding torch is proportional to the difference in the remaining amount, and the proportionality coefficient is 0.05 to 0.1 mm per square millimeter, that is, for every increase of 1 square millimeter in the difference in the remaining amount, the offset distance increases by 0.05 to 0.1 mm. The minimum value of the offset distance is 0.1 mm, and the maximum value is 3 mm. When the calculated offset distance exceeds 3 mm, it is carried out according to 3 mm, and a warning signal is issued. When the offset distance is greater than 1 mm, the welding speed is reduced by 10% to ensure good fusion on the offset side. When the difference between the remaining amount on the nozzle side and the remaining amount on the shell side is less than 0.5 square millimeter, the welding torch welds in the center without offset. Between two adjacent offset positions, the offset amount of the welding torch is smoothly transitioned in a linear interpolation manner, and the interpolation step size is not greater than 0.1 degree. After each filling layer is completed, the surface quality of the weld bead is inspected. It is qualified if there are no defects such as cracks, pores, and slag inclusions. If there are defects, the defects are removed and the layer is welded again. Repeat the processes of interlayer scanning, remaining amount calculation, and filling layer welding until the remaining amount meets the requirements of the cover layer.

[0066] For example, the remaining amount on the nozzle side is 15 square millimeters, the remaining amount on the shell side is 5 square millimeters, the difference is 10 square millimeters, and the proportionality coefficient is taken as 0.08 mm per square millimeter, then the offset distance is 0.8 mm, and the welding torch is offset 0.8 mm towards the side wall line of the nozzle. The height of the filling layer weld bead is measured with a height gauge, and 8 measurement points are evenly selected circumferentially for each weld bead layer, and the average value is taken.

[0067] The S704 performs offset welding of the cap layer, ensuring that the weld toes of the cap layer fall within the preset limits of the nozzle sidewall line and the cylinder sidewall line, respectively. The welding parameters for the cap layer are: current 120 to 180 amps, voltage 22 to 28 volts, and welding speed 120 to 180 mm per minute. The number of weld passes in the cap layer is determined by the bevel width; one pass is welded when the bevel width is less than or equal to 20 mm, and two passes are welded when it is greater than 20 mm. The overlap width between cap layer weld passes is 40% to 50% of the weld pass width to ensure a smooth cap surface. The offset and oscillation range of the welding torch are controlled to ensure that the two weld toes of the cap layer fall within the preset limits of the nozzle sidewall line and the cylinder sidewall line, respectively. The preset limits are: the maximum distance the weld toe extends beyond the sidewall line should not exceed 2 mm, and the maximum distance it extends below the sidewall line should not exceed 1 mm. The excess height of the cap crown is 0 to 3 mm, with a uniformity of no more than 1 mm. After the cap layer welding is completed, the welding torch automatically returns to the safe position. The controller initiates the cooling timer program, waiting for the weld to cool to room temperature. The cooling time should be no less than 24 hours, or the difference between the weld temperature and the ambient temperature should not exceed 5 degrees Celsius, as measured by a thermometer. Visually inspect the cover weld; it should be free of cracks, porosity, undercut, and other defects. Undercut depth should not exceed 0.5 mm, and the continuous length should not exceed 100 mm to be considered acceptable. If the cover weld has defects exceeding the acceptable limits, remove the defects and re-weld the cover layer.

[0068] The excess height of the weld cover is measured using a weld inspection gauge. Twelve measurement points are evenly selected around the weld circumference, and the maximum value is taken as the excess height. The undercut depth is measured using a depth gauge. Measurements are taken every 10 millimeters in the continuous undercut area, and the maximum value is taken as the undercut depth.

[0069] After the offset welding of the S801 cap layer is completed and cooled, the contours of the final weld surface, the outer surface of the nozzle, and the outer surface of the cylinder are acquired along the intersection trajectory to obtain the final inspection point cloud. The motion speed, scanning distance, and cross-sectional acquisition interval of the final inspection scan are completely consistent with the pre-weld scan to ensure data comparability. The final inspection scan covers the entire surface of the final weld, the outer surface of the nozzle 50 mm on each side of the weld, and the outer surface of the cylinder, facilitating the inspection of the transition between the weld toe and the base material. The ambient light intensity during scanning is no greater than 1000 lux to avoid strong light reflection affecting scanning accuracy. The scanning head performs a 360-degree continuous scan along the intersection trajectory, and position calibration is performed at each circumferential mark during the scan. The coordinate accuracy of the final inspection point cloud is no greater than 0.05 mm. If the quality of the final inspection point cloud is unsatisfactory, the final inspection scan is repeated.

[0070] Illuminance was measured using a lux meter, with measurements taken once at each of the four corners and the center of the scanned area, and the average value was recorded. During the final point cloud integrity check, the number of point clouds at all circumferential locations was counted, and each cross-section was considered complete if it contained no fewer than 100 point clouds.

[0071] The S802 module projects the final inspection point cloud onto the corresponding circumferential cross-sections, extracting the weld toe lines on the nozzle side, cylinder side, and cap crown contour. The final inspection point cloud is registered with the pre-weld point cloud, with a registration accuracy of no more than 0.05 mm. Then, the final inspection point cloud is projected onto each pre-weld cross-section to obtain the two-dimensional contour of the final weld within each cross-section. The weld toe line is the boundary line between the weld surface point cloud and the base material surface point cloud, i.e., the line connecting the points with the largest gradient change rate, extracted using a gradient-based edge detection algorithm. The cap crown contour is the highest contour line at the top of the weld; the highest point of the cap crown contour within each cross-section is the maximum Z-coordinate point of the weld surface point cloud within that cross-section, extracted using an extreme point extraction algorithm. Outliers with deviations exceeding three times the standard deviation are deleted, and then interpolation between adjacent points is used for completion. The coordinate error of the extracted contour is no more than 0.1 mm.

[0072] The gradient change rate is the height difference between adjacent points divided by the horizontal distance, where the horizontal distance is 0.1 mm. For example, in outlier handling, if a point has a height of 10 mm, and the average height of its 10 surrounding points is 5 mm with a standard deviation of 1 mm, then this point's deviation is 5 times the standard deviation, making it an outlier. After deletion, it is replaced by the average of the two surrounding points.

[0073] S803 uses the circumferential position as the horizontal axis to unfold the positions of the weld toe lines on the nozzle side, cylinder side, and cap crown relative to the initial allowance line into a two-dimensional image. A two-dimensional coordinate system is established, with the circumferential angle as the horizontal axis (scale interval 10 degrees) and the vertical distance within the cross-section as the vertical axis (scale interval 1 millimeter). The vertical distance is based on the initial allowance line, with upward values ​​being positive and downward values ​​negative, in millimeters. The deviation value equals the actual contour coordinates minus the coordinates of the corresponding position on the initial allowance line; a positive value indicates the actual contour is higher than the initial allowance line, and a negative value indicates it is lower. The deviation values ​​for the weld toe lines on the nozzle side, cylinder side, and cap crown at each circumferential position are plotted in the two-dimensional coordinate system, forming three continuous curves. Different deviation ranges are indicated by different colors: deviations within ±0.5 millimeters are represented by green, ±0.5 to 1 millimeter by yellow, and deviations exceeding ±1 millimeter by red. The resolution of the 2D image should be no less than 300 dpi. The horizontal length is determined based on the circumferential angle range, and the vertical length is determined based on the maximum deviation value.

[0074] The origin of the 2D image coordinates is set at 0 degrees, and the initial allowance line height is 0. The color coding indicates that green means the deviation is within acceptable limits, yellow means the deviation is close to the upper limit and requires attention, and red means the deviation exceeds the standard and requires rework.

[0075] S804 retains the identifiers of the four circumferential marks, generating and outputting a forming deviation diagram. In the generated 2D image, the circumferential angle positions corresponding to the four circumferential marks are marked, and the shape identifiers of the four circumferential marks are retained. The maximum positive deviation, maximum negative deviation, and their corresponding circumferential positions are marked, along with information such as welding date, workpiece number, and operator. The position of the initial allowance line is represented by a dashed line, the weld toe line on the nozzle side by a solid line, the weld toe line on the cylinder side by a dotted-dash line, and the crown outline of the cover by a double-dash line. The forming deviation diagram is saved in image and document formats, and a deviation data text file is also output. The text file contains four data items: circumferential angle, weld toe deviation on the nozzle side, weld toe deviation on the cylinder side, and crown deviation on the cover, with a data interval of 1 degree. All relevant data, including point clouds, trajectories, and deviation diagrams, are archived and stored for a period not less than the design service life of the pressure vessel. The forming deviation is judged to be acceptable according to the pressure vessel welding standards; all deviation values ​​are considered acceptable if they are within the allowable range. Once the result is satisfactory, the process ends. If the result is unsatisfactory, the location that exceeds the standard is marked and rework is required. After rework, a final inspection scan and deviation assessment are performed again.

[0076] Archived data includes pre-welding point clouds, point clouds between each layer, final inspection point clouds, all welding trajectory files, real-time welding parameter records, forming deviation diagrams, deviation data text, and welding logs. After rework, all steps from S601 to S804 are re-executed until the forming deviation is acceptable.

[0077] like Figure 1 As shown, the cylinder and the nozzle form an intersecting weld area at their intersection. The nozzle has circumferential markings indicating positions at 0 degrees, 90 degrees, 180 degrees, and 270 degrees. The controller is connected to both the scanning head and the welding actuator. The scanning head acquires the contour of the intersection between the nozzle and the cylinder along the intersecting trajectory, while the welding torch of the welding actuator performs welding along the intersecting area according to the welding trajectory generated by the controller.

[0078] The embodiments of this example have been described above. However, this example is not limited to the specific implementation methods described above. The specific implementation methods described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms based on the guidance of this example, and all of them are within the protection scope of this example.

Claims

1. A welding quality control system based on pressure vessel manufacturing, applied to welding quality control operations including a scanning head, a welding actuator with a welding torch, and a controller, for welding beveled nozzles and cylinders, characterized in that, include: The connecting pipe is clamped to the cylinder, and four circumferential marks are set outside the welding heat-affected zone to establish a mapping relationship between the circumferential position and the spatial point of the weld. Three-dimensional contour acquisition is performed along the intersecting trajectory defined by the four circumferential marks to obtain the pre-welding point cloud; Based on the pre-welding point cloud, cross sections are established at each circumferential position. The normal direction of the nozzle and the normal direction of the cylinder are extracted. The normal direction of the nozzle and the normal direction of the cylinder are used to generate the normal orientation, and the initial allowance line and the base trajectory are constructed. The welding torch of the welding actuator is controlled to perform controlled welding of the root pass and transition layer along the root pass trajectory in the normal orientation, forming a weld bead; When the weld bead cools to the interlayer temperature, the scanning head is used to collect the contour again along the intersection trajectory to obtain the interlayer point cloud, and the interlayer point cloud is projected onto the cross-section to generate the actual allowance line. Using the actual allowance line as the filling basis, extract the remaining amount on the pipe side and the remaining amount on the cylinder side within the cross-section, and control the welding torch to be offset according to the difference between the remaining amount on the pipe side and the remaining amount on the cylinder side, and perform offset welding of the remaining filling layer and the cover layer. A final inspection scan is performed on the final weld after the cover layer is completed, and the final weld contour is projected onto the cross-section to output a forming deviation diagram.

2. The welding quality control system based on pressure vessel manufacturing according to claim 1, characterized in that, The step of clamping the nozzle to the cylinder and setting four circumferential marks outside the weld heat-affected zone to establish a mapping relationship between the circumferential position and the spatial points of the weld includes: Fix the pre-drilled cylinder onto a roller frame or welding positioner; Insert the connecting pipe into the opening of the cylinder or place it on the seat on the outer surface of the cylinder; With the projection direction of the axis of the connecting pipe onto the outer surface of the cylinder as the 0-degree position, the 90-degree, 180-degree, and 270-degree positions are sequentially formed along the circumference of the connecting pipe. Four circumferential marks are made on the outer surface of the cylinder and the outer surface of the connecting pipe corresponding to the 0-degree position, the 90-degree position, the 180-degree position and the 270-degree position, respectively; The controller reads the spatial coordinates of the four circumferential markers and establishes the mapping relationship.

3. A welding quality control system based on pressure vessel manufacturing according to claim 2, characterized in that, The step of acquiring a three-dimensional contour along the intersecting trajectory defined by the four circumferential marks to obtain a pre-welding point cloud includes: The bevel at the intersection of the connector and the cylinder is cleaned. Move the scanning head to the circumferential mark corresponding to the 0-degree position, and read the relative position of the circumferential mark and the bevel edge; The scanning head is controlled to sequentially pass through the 90-degree position, the 180-degree position, and the 270-degree position along the circumference of the connecting pipe and return to the 0-degree position, thus completing 360-degree continuous acquisition; The cross-sectional profiles of the outer wall of the nozzle, the root of the bevel, and the outer wall of the cylinder are collected at various circumferential locations. The collected point cloud is used as the pre-welding point cloud.

4. A welding quality control system based on pressure vessel manufacturing according to claim 3, characterized in that, The process of establishing cross-sections at various circumferential positions based on the pre-welding point cloud, extracting the nozzle normal and cylinder normal, generating a normal orientation from the nozzle normal and cylinder normal, and constructing an initial allowance line and generating a base trajectory includes: In the pre-welding point cloud, the outer cylindrical surface of the cylinder and the outer cylindrical surface of the nozzle are fitted respectively, and the cylinder normal and the nozzle normal at the corresponding circumferential position are extracted. Extract the weld tangent along the center region of the weld and establish the cross section perpendicular to the weld tangent; Within the cross-section, the side of the bevel closest to the nozzle is defined as the nozzle sidewall line, the side of the bevel closest to the cylinder is defined as the cylinder sidewall line, and the root outline is also defined. Using the pipe sidewall line and the cylinder sidewall line as boundaries, the area to be filled in the cross section is decomposed into initial pipe side margin, initial root margin, and initial cylinder side margin. The initial nozzle side allowance, the initial root allowance, and the initial cylinder side allowance are connected in circumferential order to form the initial allowance line; The angle bisector between the normal of the nozzle and the normal of the cylinder is extracted as the normal orientation. The base trajectory is generated based on the geometric center of the initial root margin and the normal orientation.

5. A welding quality control system based on pressure vessel manufacturing according to claim 4, characterized in that, The welding torch controlling the welding actuator performs controlled welding of the root pass and transition layer along the root pass trajectory in the normal orientation, forming a weld bead, including: Move the welding torch to the preset arc-starting point so that the axis of the welding torch enters the normal posture of the corresponding circumferential position; The electric arc is initiated and advanced along the intersecting trajectory. The corresponding initial allowance line is read, and the lateral position of the welding torch is restricted within the initial root allowance to perform root pass welding. When passing the four circumferential marks, the welding torch is rotated synchronously with the normal of the cylinder and the normal of the nozzle within the cross-section. After the initial welding is completed and the arc is closed, the controlled welding of the transition layer is performed in the same coordinate system. The center line of the transition layer is obtained by translating the initial allowance line to the edge of the bevel, and the endpoints of the transition layer are kept within the range of the pipe side wall line and the cylinder side wall line.

6. A welding quality control system based on pressure vessel manufacturing according to claim 5, characterized in that, When the weld bead cools to the interlayer temperature, the scanning head is used to collect the contour again along the intersection trajectory to obtain the interlayer point cloud. The interlayer point cloud is then projected onto the cross-section to generate the actual allowance line, including: After the controlled welding of the root pass and the transition layer is completed and the weld bead cools to the interlayer temperature, the surface of the weld bead is cleaned. Move the scanning head to the 0-degree position and read the spatial coordinates of the four circumferential marks again; The surface of the weld bead, the remaining area of ​​the bevel on the pipe side, and the remaining area of ​​the bevel on the cylinder side are collected along the intersection trajectory as the interlayer point cloud; The interlayer point cloud is projected onto the cross-section to extract the actual weld lines at each circumferential position; The actual weld line is used to replace the original contour at the corresponding position in the initial allowance line to generate the actual allowance line.

7. A welding quality control system based on pressure vessel manufacturing according to claim 6, characterized in that, The process of using the actual allowance line as a filling basis, extracting the remaining amount on the nozzle side and the remaining amount on the cylinder side within the cross-section, and controlling the welding torch to be offset according to the difference between the remaining amount on the nozzle side and the remaining amount on the cylinder side, and performing offset welding of the remaining filler layer and the cover layer includes: Within the cross-section at each circumferential position, the values ​​of the remaining amount on the pipe side and the remaining amount on the cylinder side are calculated based on the actual allowance line; For each remaining fill layer, determine the relationship between the remaining amount on the pipe side and the remaining amount on the cylinder side, generate a welding trajectory close to the corresponding side wall line on the side with the larger value, and then generate an overlap trajectory towards the middle of the cross-section. The remaining filler layer is offset welded in a circumferential sequence. When the remaining amount on the pipe side is greater than the remaining amount on the cylinder side, the welding torch is controlled to offset towards the pipe side wall line; when the remaining amount on the cylinder side is greater than the remaining amount on the pipe side, the welding torch is controlled to offset towards the cylinder side wall line. Perform offset welding on the cover layer so that the weld toe of the cover layer falls within the preset limits of the pipe sidewall line and the cylinder sidewall line, respectively.

8. A welding quality control system based on pressure vessel manufacturing according to claim 7, characterized in that, The final inspection scan of the weld after the completion of the cover layer, projecting the final weld contour onto the cross-section, and outputting a forming deviation diagram includes: After the offset welding of the cover layer is completed and cooled, the contours of the surface of the final weld, the outer surface of the nozzle, and the outer surface of the cylinder are collected along the intersection trajectory to obtain the final inspection point cloud; The final inspection point cloud is projected onto the cross section at the corresponding circumferential position, and the weld toe line on the pipe side, the weld toe line on the cylinder side, and the crown contour of the cover are extracted respectively. Using the circumferential position as the horizontal axis, the positions of the weld toe line on the nozzle side, the weld toe line on the cylinder side, and the crown outline of the cover surface relative to the initial allowance line are unfolded into a two-dimensional image; The identifiers of the four circumferential marks are retained, and the forming deviation diagram is generated and output.

Images