Welding robot welding seam positioning method for shipyard part machining

Abstract

The invention discloses a welding robot welding seam positioning method for shipyard part machining. The problems that an existing method is poor in deformation adaptation, weak in interference resistance and insufficient in positioning precision for shipyard large-size parts are solved. The method comprises the four steps that 1, a to-be-welded part is pre-fixed through a shipyard special tool, and a laser contourgraph, a millimeter wave radar and an industrial camera are combined to collect multi-source features; 2, obtaining a final welding seam positioning coordinate through multi-source feature layered extraction (a global reference feature layer builds a coordinate network, and a local welding seam feature layer improves a groove / gap parameter), global-local correlation matching (a rigid body transforms a mapping coordinate and calculates a deformation deviation) and dynamic deviation compensation (cubic spline interpolation fits a deformation curve and generates a three-dimensional compensation amount); 3, a positioning coordinate calibration robot TCP is imported, and a path is generated according to the welding seam type; and 4, verifying the precision in a no-load manner, and recalculating the compensation amount if the deviation exceeds 0.1 mm. The method is high in positioning precision, strong in interference resistance and suitable for complex working conditions of shipyards.

Description

Technical Field

[0001] This invention relates to the field of welding robot positioning, specifically a method for positioning weld seams in a welding robot used in shipyard component processing. Background Technology

[0002] Shipyard components (such as butt welds on hull sections and fillet welds between ribs and outer plates) are characterized by large dimensions (single section lengths can exceed 20m), complex weld types (grooving welds, fillet welds, lap welds), easy workpiece deformation (warping caused by pre-welding material cutting errors and assembly stress), and strong environmental interference (workshop dust, welding spatter, oxide scale). Existing welding robot weld positioning methods have significant drawbacks: a disconnect between global and local positioning; reliance on pre-set CAD models for global matching, but actual workpiece deformation (e.g., deflection reaching 5-10m) is a concern. m) This leads to large deviations between the model and the actual object, and the accuracy requirements of local welds (such as bevel gaps of 0.5-2mm) cannot be met; inefficient multi-sensor data fusion: simply superimposing laser, vision, and radar data without distinguishing between "global reference features" and "local weld features" makes it susceptible to interference from dust and oxide scale, resulting in weld edge extraction errors exceeding 0.3mm; lack of dynamic deviation compensation: failing to consider changes in workpiece assembly gaps and local stress deformation before welding, resulting in weld paths deviating from the actual values ​​by more than 0.5mm after positioning, leading to defects such as weld misalignment and incomplete fusion.

[0003] Existing technologies cannot solve the coupling problem of "large-size global positioning - local weld accuracy - dynamic deformation compensation", and there is an urgent need for a special positioning method adapted to the characteristics of shipyard components. Summary of the Invention

[0004] The purpose of this invention is to provide a method for positioning weld seams in a welding robot used for shipyard component processing, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for positioning weld seams using a welding robot for shipyard component processing, comprising the following steps:

[0006] Step 1, Pre-positioning of workpiece and multi-source feature acquisition: The parts to be welded are initially fixed using shipyard-specific tooling; then, a combination of laser profilometer, millimeter-wave radar and industrial camera is used to acquire data and obtain different features in a targeted manner;

[0007] Step 2, Multi-source Feature Hierarchical Association - Dynamic Deviation Compensation: Through a three-order logic of hierarchical extraction, association matching, and dynamic compensation, the global positioning of the component to be welded and the local weld seam accuracy matching are achieved. The specific steps are as follows:

[0008] Step 2.1, Multi-source feature hierarchical extraction and filtering:

[0009] Global reference feature layer extraction: Based on millimeter-wave radar and industrial camera data, a global coordinate system is constructed, with the welding robot base coordinate system as the reference: the center coordinates P of the positioning hole are extracted. g (X g ,Y g Z g ), P g Specifically, the distance L and angle θ from the edge of the positioning hole to the radar are measured by millimeter-wave radar, and the offset Δx and Δy of the QR code center extracted by the camera are calculated. Valid reference features are screened: positioning holes with a ranging error >2mm due to tooling obstruction are removed, and ≥3 valid feature points are retained to construct a global coordinate network.

[0010] Local weld feature layer extraction: Based on laser profilometer data, focusing on the core parameters of the weld and eliminating interfering features; Groove profile fitting: Performing polynomial fitting on the weld cross-sectional data acquired by the laser to extract the upper edge A(X) of the groove. A ,Y A Z A ), lower edge B(X) B ,Y B Z B ), calculate the bevel angle Assembly clearance calculation: Let the root edge point be A′(X) A ′,Y A ′,Z A B′(X′) and B′(X′) B ,Y B ′,Z′ B Extract the weld root gap d = |X A ′-X′ B | Filter effective weld segments with d in the range of 0.3-2.5mm; Feature denoising: remove contour abrupt changes caused by oxide scale, and use moving average filtering to smooth the contour curve;

[0011] Step 2.2, Global-Local Feature Association Matching:

[0012] Coordinate mapping model establishment: Using the positioning hole coordinates of the global reference feature layer as the origin, establish a global coordinate system O-XTZ, and map the laser coordinate system O′-x′y′z′ of the local weld features to the global coordinate system through rigid body transformation;

[0013] Deformation deviation calculation: Compare the actual coordinates of the mapped weld with the design coordinates to calculate the deviation ΔP(X) caused by workpiece deformation. Δ ,Y Δ Z Δ ):

[0014]

[0015] If ΔP≤0.2mm, proceed directly to step 3; if ΔP>0.2mm, activate dynamic deviation compensation.

[0016] Step 2.3, Dynamic Deviation Compensation:

[0017] Deformation trend analysis: A cross-section was collected every 50 mm along the weld length using a laser profilometer. The deformation amount ΔZ1 (i=1,2,…,n) at the height of each cross-section was calculated. The deformation curve Z(x)=ax was fitted using cubic spline interpolation. 7 +bx 2 +cx+d, where x is the coordinate of the weld length direction; a, b, c are fitting coefficients, constants obtained by fitting actual deformation sampling data;

[0018] Compensation parameter generation: Calculate the compensation amount ΔZ for each weld point based on the deformation curve. 补偿 ΔZ 补偿 =Z(x)-Z 设计 (x), where Z 设计 (x) represents the weld design height curve; similarly, ΔX is obtained. 补偿 =X 变形 (x)-X 设计 (x), ΔY 补偿 =Y 变形 (y)-Y 设计 (y), in the above formula, X 变形 (x) and Y 变形 (y) represents the actual deformation values ​​of the weld at coordinates x in the length direction and y in the width direction, respectively. The actual deformation value is the deviation between the mapped actual coordinates and the pre-positioned coordinates; X 设计 (x) and Y 设计 (y) represents the design values ​​of the weld at coordinate x in the length direction and coordinate y in the width direction. The design values ​​are the theoretical coordinates marked on the drawings.

[0019] Positioning coordinate correction: The compensation amount is superimposed on the original mapped coordinates to obtain the final weld positioning coordinates: P 最终 (X 实际 +ΔX 补偿 ,Y 实际 +ΔY 补偿 Z 实际 +ΔZ 补偿 Ensure that ΔP ≤ 0.1 mm after correction;

[0020] Step 3, Weld Coordinate Calibration and Path Generation: Import the final weld positioning coordinates into the welding robot control system, with the weld start point P: and end point P: as the coordinates. ； Inflection point P <For key nodes, calibrate the robot TCP, i.e., the relative position of the tool center point and the weld; according to the weld type, generate the robot welding path based on the positioning coordinates, with the path node spacing ≤10mm to ensure smooth movement;

[0021] Step 4, Positioning Accuracy Verification: Start the robot to run without load, and measure the distance between the TCP and the weld edge in real time using the laser displacement sensor installed at the end, and record the maximum deviation value; if the maximum deviation is >0.1mm, return to step 2.3, increase the density of deformation sampling points, and recalculate the compensation amount; if the deviation is ≤0.1mm, positioning is complete, and welding operation is started.

[0022] Preferably, in step 1, the deviation between the global coordinates of the component to be welded initially fixed and the coordinates of the welding robot base is controlled within ±5mm; at the same time, the oxide scale and oil stains within 100mm around the weld are cleaned to expose the positioning marks.

[0023] Preferably, in step 2.1, the center coordinate P of the positioning hole is... g (X g ,Y g Z g The formula for calculating X is: g =L×cosθ+Δx,Y g =L×sinθ+Δy,Z g =H A +Δz, where H A Δz represents the design height of the positioning hole, and Δz represents the deviation between the actual height of the positioning hole measured by the laser profilometer and the design value.

[0024] Preferably, the calculation formula for mapping the laser coordinate system O′-x′y′z′ of the local weld feature to the global coordinate system through rigid body transformation in step 2.2 is as follows:

[0025]

[0026] Where R is a 3×3 rotation matrix, calculated from the design angle between the positioning hole and the weld; T is a translation vector, specifically determined by the design distance from the center of the positioning hole to the starting point of the weld.

[0027] Preferably, the specific implementation steps for weld coordinate calibration and path generation in step 3 are as follows:

[0028] Step 3.1, Weld coordinate calibration:

[0029] Step 3.1.1, Importing and Mapping Key Node Coordinates: From the final weld positioning coordinates output in Step 2.3, select three types of core nodes:

[0030] Weld start point P: (X: end, Y: end, Z: end): The starting end of the weld; Weld end point P； (X ； Finally, Y ； Finally, Z ； (End): The end of the weld, with a 5mm gap between it and the adjacent weld; weld inflection point P < (X <终 ,Y <终 Z <终 For non-straight seam welds, extract key inflection points including but not limited to the center of the arc and corner points. For arc welds, record the center coordinates O(X). F ,Y F Z F ) and radius R;

[0031] Step 3.1.2, Mapping Coordinates to the Robot System: The "global coordinates" from step 3.1.1, specifically using the robot base as a reference, are mapped to the "tool coordinate system" of the TCP through the coordinate transformation function of the robot control system. The specific transformation formula is as follows:

[0032]

[0033] Among them, R HIK T is the rotation matrix of the TCP tool coordinate system relative to the robot base (determined by the welding torch mounting angle, such as a 45° angle between the welding torch and the weld seam during fillet welding). HIK The translation vector from the TCP origin to the welding torch nozzle (determined by the welding torch model, such as 150mm in length) ensures that the deviation between the mapped TCP coordinates and the actual weld position is ≤0.05mm.

[0034] Step 3.1.3, Relative position calibration of TCP and weld: Control the robot TCP to move unloaded to 10mm directly above the weld start point P, activate the laser displacement sensor (installed next to TCP), and measure the actual distance D between TCP and the upper edge of the weld bevel. 实测 ; Compare with D 实测 Distance D from the design 设计 (For example, if the design distance is 5mm for fillet welds), if the deviation is >0.05mm, manually fine-tune the position of TCP on the X / Y / Z axes until the deviation is ≤0.05mm; repeat step 3.1.3 to calibrate the weld endpoint P. ； With inflection point P < This ensures that the TCP positioning deviation of the three key nodes is ≤0.05mm, forming a three-point calibration benchmark to avoid global deviation caused by calibration of a single node;

[0035] Step 3.1.4 and Stage 3.1.3: Coordinate System Fixation and Saving. This involves fixing the relative position parameters of the calibrated TCP and the weld, i.e., the rotation matrix R. HIK Translation vector T HIKThe TCP coordinates of key nodes are saved to the "weld positioning parameter library" of the robot control system and associated with the current part ID (such as "rib-001"), so that they can be directly loaded when calling the same type of weld in the future, reducing the time of repeated calibration.

[0036] Step 3.2, Weld Path Generation:

[0037] Step 3.2.1, Weld Type Identification and Path Rule Matching: Based on the local weld features (groove angle, cross-sectional shape) extracted in Step 2.1, the type is automatically determined—fillet weld (groove 90°, such as rib to outer plate connection), butt weld (groove 60°-80°, such as segmented butt joint), and circular arc weld (curved trajectory, such as bulkhead ring connection); the corresponding path rules are matched, and the core parameters are preset (fillet weld "zigzag" swing amplitude 3-5mm, butt weld straight swing 1-2mm, circular arc weld node spacing ≤5mm) to ensure that the path adapts to the weld forming requirements;

[0038] Step 3.2.2, Path Node Interpolation Generation: For straight welds (fillet welds, butt welds), use the calibrated start point P: and end point P: ； Based on this, interpolate evenly at intervals of ≤10mm to generate intermediate nodes (e.g., 11 nodes for a 100mm long weld), ensuring the trajectory fits the weld centerline; Curve segment generation: For circular arc welds, based on inflection point P... < The center coordinates and radius of the circle are interpolated with an arc length ≤ 5mm, and the coordinates of each node are calculated using trigonometric functions to ensure a smooth arc.

[0039] Step 3.2.3, Path Smoothing and Collision Optimization: For adjacent nodes in Step 3.2.2, a cubic polynomial transition is used to avoid sudden changes in robot speed that cause welding torch jitter; import the 3D model of the component and detect the path and tooling distance.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] This invention uses "multi-source feature hierarchical association - dynamic deviation compensation" combined with cubic spline interpolation to fit the deformation curve, accurately correcting the three-dimensional deformation of the workpiece caused by assembly stress and material cutting error. The positioning deviation is reduced from the traditional 0.5-1mm to ≤0.1mm, meeting the high-precision welding requirements of large-size components such as hull sections and ribs.

[0042] Anti-interference capability adapted to shipyard environment: It distinguishes between global reference (millimeter-wave radar anti-dust, industrial camera marking) and local weld features (laser profilometer bevel measurement), avoiding interference from workshop dust and oxide scale on a single sensor, with weld edge extraction error ≤0.05mm, and significantly improved positioning stability.

[0043] Improve welding efficiency and quality: The weld path generation is adapted to fillet welds, butt welds, and arc welds, with node spacing ≤10mm and smooth transition. The robot does not require manual reprogramming, and the positioning time for a single weld is reduced from 15-20 minutes to 3-5 minutes. At the same time, it reduces weld deviation and incomplete fusion defects, and improves the welding qualification rate by 15%-20%.

[0044] Reduce the cost of repetitive operations: Calibration parameters and path are associated with component IDs and archived, and similar welds can be directly called, reducing the time for repeated calibration; the no-load accuracy verification process avoids rework after welding, further reducing the time and material costs of shipyard component processing. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0046] Figure 2 This is a schematic diagram of the multi-source feature hierarchical association-dynamic deviation compensation method of the present invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Please see Figure 1-2 This invention provides a technical solution: a method for positioning weld seams using a welding robot for shipyard component processing, comprising the following steps:

[0049] Step 1, Pre-positioning of workpiece and multi-source feature acquisition: The parts to be welded are initially fixed using shipyard-specific tooling; then, a combination of laser profilometer, millimeter-wave radar and industrial camera is used to acquire data and obtain different features in a targeted manner;

[0050] Step 2, Multi-source Feature Hierarchical Association - Dynamic Deviation Compensation: Through a three-order logic of hierarchical extraction, association matching, and dynamic compensation, the global positioning of the component to be welded and the local weld seam accuracy matching are achieved. The specific steps are as follows:

[0051] Step 2.1, Multi-source feature hierarchical extraction and filtering:

[0052] Global reference feature layer extraction: Based on millimeter-wave radar and industrial camera data, a global coordinate system is constructed, with the welding robot base coordinate system as the reference: the center coordinates P of the positioning hole are extracted. g (X g ,Y g Z g ), Pg Specifically, the distance L and angle θ from the edge of the positioning hole to the radar are measured by millimeter-wave radar, and the offset Δx and Δy of the QR code center extracted by the camera are calculated. Valid reference features are screened: positioning holes with a ranging error >2mm due to tooling obstruction are removed, and ≥3 valid feature points are retained to construct a global coordinate network.

[0053] Local weld feature layer extraction: Based on laser profilometer data, focusing on the core parameters of the weld and eliminating interfering features; Groove profile fitting: Performing polynomial fitting on the weld cross-sectional data acquired by the laser to extract the upper edge A(X) of the groove. A ,Y A Z A ), lower edge B(X) B ,Y B Z B ), calculate the bevel angle Assembly clearance calculation: Let the root edge point be A′(X) A ′,Y A ′,Z A B′(X′) and B′(X′) B ,Y B ′,Z′ B Extract the weld root gap d = |X A ′-X′ B | Filter effective weld segments with d in the range of 0.3-2.5mm; Feature denoising: remove contour abrupt changes caused by oxide scale, and use moving average filtering to smooth the contour curve;

[0054] Step 2.2, Global-Local Feature Association Matching:

[0055] Coordinate mapping model establishment: Using the positioning hole coordinates of the global reference feature layer as the origin, establish a global coordinate system O-XYZ, and map the laser coordinate system O′-x′y′z′ of the local weld features to the global coordinate system through rigid body transformation;

[0056] Deformation deviation calculation: Compare the actual coordinates of the mapped weld with the design coordinates to calculate the deviation ΔP(X) caused by workpiece deformation. Δ ,Y Δ Z Δ ):

[0057]

[0058] If ΔP≤0.2mm, proceed directly to step 3; if ΔP>0.2mm, activate dynamic deviation compensation.

[0059] Step 2.3, Dynamic Deviation Compensation:

[0060] Deformation trend analysis: A cross-section was collected every 50 mm along the weld length using a laser profilometer. The deformation amount ΔZ1 (i=1,2,…,n) at the height of each cross-section was calculated. The deformation curve Z(x)=ax was fitted using cubic spline interpolation. 7 +bx 2 +cx+d, where x is the coordinate of the weld length direction; a, b, c are fitting coefficients, constants obtained by fitting actual deformation sampling data;

[0061] Compensation parameter generation: Calculate the compensation amount ΔZ for each weld point based on the deformation curve. 补偿 ΔZ 补偿 =Z(x)-Z 设计 (x), where Z 设计 (x) represents the weld design height curve; similarly, ΔX is obtained. 补偿 =X 变形 (x)-X 设计 (x), ΔY 补偿 =Y 变形 (y)-Y 设计 (y), in the above formula, X 变形 (x) and Y 变形 (y) represents the actual deformation values ​​of the weld at coordinates x in the length direction and y in the width direction, respectively. The actual deformation value is the deviation between the mapped actual coordinates and the pre-positioned coordinates; X 设计 (x) and Y 设计 (y) represents the design values ​​of the weld at coordinate x in the length direction and coordinate y in the width direction. The design values ​​are the theoretical coordinates marked on the drawings.

[0062] Positioning coordinate correction: The compensation amount is superimposed on the original mapped coordinates to obtain the final weld positioning coordinates: P 最终 (X 实际 +ΔX 补偿 ,Y 实际 +ΔY 补偿 Z 实际 +ΔZ 补偿 Ensure that ΔP ≤ 0.1 mm after correction;

[0063] Step 3, Weld Coordinate Calibration and Path Generation: Import the final weld positioning coordinates into the welding robot control system, with the weld start point P: and end point P: as the coordinates. ； Inflection point P < For key nodes, calibrate the robot TCP, i.e., the relative position of the tool center point and the weld; according to the weld type, generate the robot welding path based on the positioning coordinates, with the path node spacing ≤10mm to ensure smooth movement;

[0064] Step 4, Positioning Accuracy Verification: Start the robot to run without load, and measure the distance between the TCP and the weld edge in real time using the laser displacement sensor installed at the end, and record the maximum deviation value; if the maximum deviation is >0.1mm, return to step 2.3, increase the density of deformation sampling points, and recalculate the compensation amount; if the deviation is ≤0.1mm, positioning is complete, and welding operation is started.

[0065] Furthermore, in step 1, the deviation between the global coordinates of the component to be welded and the coordinates of the welding robot base is controlled within ±5mm; at the same time, the oxide scale and oil stains within 100mm around the weld are cleaned to expose the positioning marks.

[0066] Furthermore, in step 2.1, the center coordinate P of the positioning hole... g (X g ,Y g Z g The formula for calculating X is: g =L×cosθ+Δx,Y g =L×sinθ+Δy,Z g =H A +Δz, where H A Δz represents the design height of the positioning hole, and Δz represents the deviation between the actual height of the positioning hole measured by the laser profilometer and the design value.

[0067] Furthermore, in step 2.2, the calculation formula for mapping the laser coordinate system O′-x′y′z′ of the local weld features to the global coordinate system through rigid body transformation is as follows:

[0068]

[0069] Where R is a 3×3 rotation matrix, calculated from the design angle between the positioning hole and the weld; T is a translation vector, specifically determined by the design distance from the center of the positioning hole to the starting point of the weld.

[0070] Furthermore, the specific implementation steps for weld coordinate calibration and path generation in step 3 are as follows:

[0071] Step 3.1, Weld coordinate calibration:

[0072] Step 3.1.1, Importing and Mapping Key Node Coordinates: From the final weld positioning coordinates output in Step 2.3, select three types of core nodes:

[0073] Weld start point P: (X: end, Y: end, Z: end): The starting end of the weld; Weld end point P ； (X ； Finally, Y ； Finally, Z ； (End): The end of the weld, with a 5mm gap between it and the adjacent weld; weld inflection point P <(X <终 ,Y <终 Z <终 For non-straight seam welds, extract key inflection points including but not limited to the center of the arc and corner points. For arc welds, record the center coordinates O(X). F ,Y F Z F ) and radius R;

[0074] Step 3.1.2, Mapping Coordinates to the Robot System: The "global coordinates" from step 3.1.1, specifically using the robot base as a reference, are mapped to the "tool coordinate system" of the TCP through the coordinate transformation function of the robot control system. The specific transformation formula is as follows:

[0075]

[0076] Among them, R HIK T is the rotation matrix of the TCP tool coordinate system relative to the robot base (determined by the welding torch mounting angle, such as a 45° angle between the welding torch and the weld seam during fillet welding). HIK The translation vector from the TCP origin to the welding torch nozzle (determined by the welding torch model, such as 150mm in length) ensures that the deviation between the mapped TCP coordinates and the actual weld position is ≤0.05mm.

[0077] Step 3.1.3, Relative position calibration of TCP and weld: Control the robot TCP to move unloaded to 10mm directly above the weld start point P, activate the laser displacement sensor (installed next to TCP), and measure the actual distance D between TCP and the upper edge of the weld bevel. 实测 ; Compare with D 实测 Distance D from the design 设计 (For example, if the design distance is 5mm for fillet welds), if the deviation is >0.05mm, manually fine-tune the position of TCP on the X / Y / Z axes until the deviation is ≤0.05mm; repeat step 3.1.3 to calibrate the weld endpoint P. ； With inflection point P < This ensures that the TCP positioning deviation of the three key nodes is ≤0.05mm, forming a three-point calibration benchmark to avoid global deviation caused by calibration of a single node.

[0078] Step 3.1.4 and Stage 3.1.3: Coordinate System Fixation and Saving. This involves fixing the relative position parameters of the calibrated TCP and the weld, i.e., the rotation matrix R. HIK Translation vector T HIK The TCP coordinates of key nodes are saved to the "weld positioning parameter library" of the robot control system and associated with the current part ID (such as "rib-001"), so that they can be directly loaded when calling the same type of weld in the future, reducing the time of repeated calibration.

[0079] Step 3.2, Weld Path Generation:

[0080] Step 3.2.1, Weld Type Identification and Path Rule Matching: Based on the local weld features (groove angle, cross-sectional shape) extracted in Step 2.1, the type is automatically determined—fillet weld (groove 90°, such as rib to outer plate connection), butt weld (groove 60°-80°, such as segmented butt joint), and circular arc weld (curved trajectory, such as bulkhead ring connection); the corresponding path rules are matched, and the core parameters are preset (fillet weld "zigzag" swing amplitude 3-5mm, butt weld straight swing 1-2mm, circular arc weld node spacing ≤5mm) to ensure that the path adapts to the weld forming requirements;

[0081] Step 3.2.2, Path Node Interpolation Generation: For straight welds (fillet welds, butt welds), use the calibrated start point P: and end point P: ； Based on the baseline, interpolate evenly at intervals of ≤10mm to generate intermediate nodes (e.g., 11 nodes for a 100mm long weld), ensuring the trajectory fits the weld centerline; Curve segment generation: For circular arc welds, based on inflection point P... < The center coordinates and radius of the circle are interpolated with an arc length ≤ 5mm, and the coordinates of each node are calculated using trigonometric functions to ensure a smooth arc.

[0082] Step 3.2.3, Path Smoothing and Collision Optimization: For adjacent nodes in Step 3.2.2, a cubic polynomial transition is used to avoid sudden changes in robot speed that cause welding torch jitter; import the 3D model of the component and detect the path and tooling distance.

[0083] This invention utilizes multi-source feature hierarchical association and dynamic deviation compensation, combined with cubic spline interpolation to fit deformation curves, to accurately correct the three-dimensional deformation of large-sized components (such as hull sections and ribs) in shipyards. The positioning deviation is reduced to ≤0.1mm, solving the positioning inaccuracy problem caused by deformation in traditional methods. This invention distinguishes the functions of millimeter-wave radar (dust resistant), industrial cameras (marker identification), and laser profilometers (bevel measurement), avoiding interference from workshop dust and oxide scale on a single sensor, thus ensuring positioning stability. The TCP calibration parameters and paths of this invention can be associated with component IDs for archiving and reuse, shortening the positioning time for single weld seams; the path generation is adaptable to multiple weld seam types and smooth, reducing weld deviation defects, improving welding pass rates, and reducing rework and repetitive operation costs.

[0084] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for positioning weld seams using a welding robot for shipyard component processing, characterized in that, Includes the following steps: Step 1, Pre-positioning of workpiece and multi-source feature acquisition: The parts to be welded are initially fixed using shipyard-specific tooling; then, a combination of laser profilometer, millimeter-wave radar and industrial camera is used to acquire data and obtain different features in a targeted manner; Step 2, Multi-source Feature Hierarchical Association - Dynamic Deviation Compensation: Through a three-order logic of hierarchical extraction, association matching, and dynamic compensation, the global positioning of the component to be welded and the local weld seam accuracy matching are achieved. The specific steps are as follows: Step 2.1, Multi-source feature hierarchical extraction and filtering: Global reference feature layer extraction: Based on millimeter-wave radar and industrial camera data, a global coordinate system is constructed, with the welding robot base coordinate system as the reference: the center coordinates P of the positioning hole are extracted. g (X g ,Y g Z g ), P g Specifically, the distance L and angle θ from the edge of the positioning hole to the radar are measured by millimeter-wave radar, and the offset Δx and Δy of the QR code center extracted by the camera are calculated. Valid reference features are screened: positioning holes with a ranging error >2mm due to tooling obstruction are removed, and ≥3 valid feature points are retained to construct a global coordinate network. Local weld feature layer extraction: Based on laser profilometer data, focusing on the core parameters of the weld and eliminating interfering features; Groove profile fitting: Performing polynomial fitting on the weld cross-sectional data acquired by the laser to extract the upper edge A(X) of the groove. A ,Y A Z A ), lower edge B(X) B ,Y B Z B ), calculate the bevel angle Assembly clearance calculation: Let the root edge point be A′(X) A ′,Y A ′,Z A B′(X′) and B′(X′) B ,Y B ′,Z′ B Extract the weld root gap d = |X A ′-X′ B | Filter effective weld segments with d in the range of 0.3-2.5mm; Feature denoising: remove contour abrupt changes caused by oxide scale, and use moving average filtering to smooth the contour curve; Step 2.2, Global-Local Feature Association Matching: Coordinate mapping model establishment: Using the positioning hole coordinates of the global reference feature layer as the origin, establish a global coordinate system O-XYZ, and map the laser coordinate system O′-x′y′z′ of the local weld features to the global coordinate system through rigid body transformation; Deformation deviation calculation: Compare the actual coordinates of the mapped weld with the design coordinates to calculate the deviation ΔP(X) caused by workpiece deformation. Δ ,Y Δ Z Δ ): If ΔP≤0.2mm, proceed directly to step 3; if ΔP>0.2mm, activate dynamic deviation compensation. Step 2.3, Dynamic Deviation Compensation: Deformation trend analysis: A cross-section was collected every 50 mm along the weld length using a laser profilometer. The deformation amount ΔZ1 (i=1,2,…,n) at the height of each cross-section was calculated. The deformation curve Z(x)=ax was fitted using cubic spline interpolation. 7 +bx 2 +cx+d, where x is the coordinate of the weld length direction; a, b, c are fitting coefficients, constants obtained by fitting actual deformation sampling data; Compensation parameter generation: Calculate the compensation amount ΔZ for each weld point based on the deformation curve. 补偿 ΔZ 补偿 =Z(x)-Z 设计 (x), where Z 设计 (x) represents the weld design height curve; similarly, ΔX is obtained. 补偿 =x 变形 (x)-X 设计 (x), ΔY 补偿 =Y 变形 (y)-Y 设计 (y), in the above formula, X 变形 (x) and Y 变形 (y) represents the actual deformation values ​​of the weld at coordinates x in the length direction and y in the width direction, respectively. The actual deformation value is the deviation between the mapped actual coordinates and the pre-positioned coordinates; X 设计 (x) and Y 设计 (y) represents the design values ​​of the weld at coordinate x in the length direction and coordinate y in the width direction. The design values ​​are the theoretical coordinates marked on the drawings. Positioning coordinate correction: The compensation amount is superimposed on the original mapped coordinates to obtain the final weld positioning coordinates: P 最终 (X 实际 +ΔX 补偿 ,Y 实际 +ΔY 补偿 Z 实际 +ΔZ 补偿 Ensure that ΔP ≤ 0.1 mm after correction; Step 3, Weld Coordinate Calibration and Path Generation: Import the final weld positioning coordinates into the welding robot control system, with the weld starting point P... s End point P ； Inflection point P < For key nodes, calibrate the robot TCP, i.e., the relative position of the tool center point and the weld; according to the weld type, generate the robot welding path based on the positioning coordinates, with the path node spacing ≤10mm to ensure smooth movement; Step 4, Positioning Accuracy Verification: Start the robot to run without load, and measure the distance between the TCP and the weld edge in real time using the laser displacement sensor installed at the end, and record the maximum deviation value; if the maximum deviation is >0.1mm, return to step 2.3, increase the density of deformation sampling points, and recalculate the compensation amount; if the deviation is ≤0.1mm, positioning is complete, and welding operation is started.

2. The method for positioning weld seams using a welding robot for shipyard component processing according to claim 1, characterized in that: In step 1, the deviation between the global coordinates of the component to be welded and the coordinates of the welding robot base is controlled within ±5mm; at the same time, the oxide scale and oil stains within 100mm around the weld are cleaned to expose the positioning marks.

3. The method for positioning weld seams using a welding robot for shipyard component processing according to claim 1, characterized in that: The center coordinate P of the positioning hole in step 2.1 g (X g ,Y g Z g The formula for calculating X is: g =L×cosθ+Δx,Y g =L×sinθ+Δy,Z g =H A +Δz, where H A Δz represents the design height of the positioning hole, and Δz represents the deviation between the actual height of the positioning hole measured by the laser profilometer and the design value.

4. The method for positioning weld seams using a welding robot for shipyard component processing according to claim 1, characterized in that: The calculation formula for mapping the laser coordinate system O′-x′y′z′ of the local weld feature to the global coordinate system through rigid body transformation in step 2.2 is as follows: Where R is a 3×3 rotation matrix, calculated from the design angle between the positioning hole and the weld; T is a translation vector, specifically determined by the design distance from the center of the positioning hole to the starting point of the weld.

5. The method for positioning weld seams using a welding robot for shipyard component processing according to claim 1, characterized in that: The specific implementation steps for weld coordinate calibration and path generation in step 3 are as follows: Step 3.1, Weld coordinate calibration: Step 3.1.1, Importing and Mapping Key Node Coordinates: From the final weld positioning coordinates output in Step 2.3, select three types of core nodes: Weld start point P: (X: end, Y: end, Z: end): The starting end of the weld; Weld end point P ； (X ； Finally, Y ； Finally, Z ； (End): The end of the weld, with a 5mm gap between it and the adjacent weld; weld inflection point P < (X <终 ,Y <终 Z <终 For non-straight seam welds, extract key inflection points including but not limited to the center of the arc and corner points. For arc welds, record the center coordinates O(X). F ,Y F Z F ) and radius R; Step 3.1.2, Mapping Coordinates to the Robot System: The "global coordinates" from step 3.1.1, specifically using the robot base as a reference, are mapped to the "tool coordinate system" of the TCP through the coordinate transformation function of the robot control system. The specific transformation formula is as follows: Among them, R HIK Let T be the rotation matrix of the TCP tool coordinate system relative to the robot base. HIK The translation vector from the TCP origin to the welding torch nozzle ensures that the deviation between the mapped TCP coordinates and the actual weld position is ≤0.05mm. Step 3.1.3, Relative position calibration of TCP and weld: Control the robot TCP to move unloaded to 10mm directly above the weld start point P, activate the laser displacement sensor (installed next to TCP), and measure the actual distance D between TCP and the upper edge of the weld bevel. 实测 ; Compare with D 实测 Distance D from the design 设计 If the deviation is >0.05mm, manually fine-tune the position of TCP on the X / Y / Z axes until the deviation is ≤0.05mm; repeat step 3.1.3 to calibrate the weld endpoint P. ； With inflection point P < This ensures that the TCP positioning deviation of the three key nodes is ≤0.05mm, forming a three-point calibration benchmark to avoid global deviation caused by calibration of a single node; Step 3.1.4 and Stage 3.1.3: Coordinate System Fixation and Saving. This involves fixing the relative position parameters of the calibrated TCP and the weld, i.e., the rotation matrix R. HIK Translation vector T HIK The TCP coordinates of key nodes are saved to the "weld positioning parameter library" of the robot control system and associated with the current component ID, so that they can be directly loaded when calling the same type of weld in the future, reducing the time of repeated calibration. Step 3.2, Weld Path Generation: Step 3.2.1, Weld type identification and path rule matching: Based on the local weld features extracted in Step 2.1, the type is automatically determined—fillet weld, butt weld, and circular arc weld; the corresponding path rules are matched, and core parameters are preset to ensure that the path adapts to the weld forming requirements; Step 3.2.2, Path Node Interpolation Generation: For straight welds, using the calibrated start point P: and end point P: ； Using this as a baseline, interpolate evenly at intervals of ≤10mm to generate intermediate nodes, ensuring the trajectory closely matches the weld centerline; Curve segment generation: For circular arc welds, based on inflection point P... < The center coordinates and radius of the circle are interpolated with an arc length ≤ 5mm, and the coordinates of each node are calculated using trigonometric functions to ensure a smooth arc. Step 3.2.3, Path Smoothing and Collision Optimization: For adjacent nodes in Step 3.2.2, a cubic polynomial transition is used to avoid sudden changes in robot speed that cause welding torch jitter; import the 3D model of the component and detect the path and tooling distance.

Images