A ship intelligent visual welding system and a welding method

Abstract

The application discloses a kind of ship intelligent visual welding systems and welding method, comprising: gantry is four column double beam structure, top is equipped with linear moving line rail that can move along Y axis direction;Movable trolley is installed on the linear moving line rail of gantry top, and the Z axis of movable trolley has rotational freedom;Welding robot is installed on movable trolley, for the weld of workpiece to be welded is welded;First vision system is arranged on gantry, for the global scanning of workpiece to be welded is generated workpiece point cloud model;Second vision system is arranged at the welding torch end of welding robot, for the accurate positioning of weld;Intelligent decision system is respectively connected with first vision system, second vision system and welding robot communication connection.The application has the advantages that the problems of traditional robot welding programming complexity, long time-consuming are solved, and the professional skill threshold and labor intensity of operator are reduced.

Description

Technical Field

[0001] This invention relates to the field of robotic welding technology, and in particular to an intelligent vision welding system and welding method for ships. Background Technology

[0002] In shipbuilding, large steel structure construction and other fields, sub-group welding is a basic and critical process. The fixed plate sub-group is a typical simple structural workpiece, mainly composed of plate material, and includes various weld forms such as flat fillet welds, vertical fillet welds and wrap fillet welds.

[0003] Currently, small-scale assembly welding operations have a low degree of automation and rely heavily on human experience. Existing small-scale assembly welding largely depends on manual operation or semi-automatic welding. Workers are exposed to welding fumes, high temperatures, strong light, and noise for extended periods, resulting in high health risks and strenuous labor. Welding quality is highly dependent on the individual skill level of the welder, making it difficult to guarantee consistency and stability. Furthermore, the industry faces a shortage and gap in skilled welders.

[0004] Traditional robot welding programming is complex and has poor adaptability. When using robots for automated welding, the traditional method requires manual teaching of the points and writing of robot programs. When the workpiece clamping position is offset, there are assembly errors in the workpiece, or the workpiece type changes, it is necessary to re-teach and reprogram, resulting in the predicament of "two hours of programming, five minutes of welding". This seriously restricts the promotion and use of robots in this field. In particular, the shipbuilding industry has typical non-standard characteristics, with many workpiece specifications and small batches, which puts forward higher requirements for the flexibility and adaptability of automated welding systems.

[0005] Existing vision positioning solutions have limitations. In some vision-guided welding solutions, sensors are usually installed near the welding torch. During the welding process, they are easily affected by arc light, spatter, high temperature and dust, which can cause sensor lenses to become contaminated, recognition failures, and reduced system reliability. Some solutions use a single vision system, which is difficult to meet the dual requirements of large-area scanning and high-precision positioning. In addition, for large workpieces (up to 20 meters in length and 10 meters in width), vision acquisition from a single station is difficult to cover completely, requiring multiple moves or splicing, which affects work efficiency.

[0006] Matching process parameters is difficult. Robotic welding requires precise welding process parameters, but traditional process databases are often based on experiments under fixed working conditions, making it difficult to exhaustively account for all changes in workpieces (such as changes in plate thickness, changes in bevel shape, etc.). As a result, when faced with new workpieces or changes in working conditions, process parameters need to be manually adjusted repeatedly. Summary of the Invention

[0007] The purpose of this invention is to provide a ship intelligent vision welding system and welding method, which solves the problems of complex and time-consuming programming in traditional robot welding, and reduces the professional skill threshold and labor intensity of operators.

[0008] The above-mentioned technical objective of the present invention is achieved through the following technical solution: A ship intelligent vision welding system includes: a gantry, a mobile trolley, a welding robot, a first vision system, a second vision system, and an intelligent decision-making system; The gantry frame is a four-column double-beam structure, with a linear moving rail on the top that can move along the Y-axis. The movable trolley is mounted on a linear moving rail on the top of the gantry, and the Z-axis of the movable trolley has a rotational degree of freedom. The welding robot is mounted on a mobile trolley and is used to weld the weld seams of the workpiece to be welded. The first vision system is mounted on a gantry and is used to perform a global scan of the workpiece to be welded and generate a point cloud model of the workpiece. The second vision system is located at the welding torch end of the welding robot and is used to accurately locate the weld seam; The intelligent decision-making system is communicatively connected to the first vision system, the second vision system, and the welding robot.

[0009] Preferably, the bottom of the gantry is provided with a ground rail, and the gantry moves along the X-axis on the ground rail. The movement of the gantry, the linear moving rail and the movable trolley are all controlled by the robot servo external axis, and its walking repeatability positioning accuracy is ±0.5mm for the X, Y and Z axes.

[0010] Preferably, the Z-axis of the movable trolley is rotatable to adjust the working posture of the correspondingly installed welding robot.

[0011] Preferably, the bottom of the welding robot is 1.6m above the ground to cover the workpiece on the ground.

[0012] Preferably, the first vision system includes a large-range lidar mounted on a gantry, used to perform an overall scan of the workpiece before welding, acquire point cloud data of the workpiece surface, and perform reverse modeling. The second vision system includes an adaptive laser mounted on the welding torch of the welding robot for precise scanning and positioning of the identified weld seam.

[0013] A smart vision welding method for ships includes the following steps: S1. Control the movement of the gantry frame, and perform a global scan of the workpiece after point-fixed assembly through the first vision system to obtain the global point cloud data of the workpiece, and reverse reconstruct the three-dimensional model of the workpiece through the intelligent decision system. S2. The intelligent decision-making system automatically identifies the plate combination in the three-dimensional model of the workpiece, extracts the weld information, and automatically generates the welding program and coarse positioning data of the weld according to the preset process database. S3. The intelligent decision-making system controls the welding robot to move to the vicinity of the weld seam based on the coarse positioning data of the weld seam. The second vision system accurately scans the arc starting point and arc ending point of the weld seam to obtain the actual spatial position and posture of the weld seam. S4. The intelligent decision-making system corrects the welding program generated in step S2 based on the actual spatial position and orientation of the weld, and controls the welding robot to perform welding operations along the corrected trajectory.

[0014] Preferably, in step S3, the precise scanning of the arc initiation point and arc extinguishing point of the weld further includes: adjusting the posture of the welding robot so that the adaptive laser is aligned with the point to be measured at a preset tilt angle and rotation angle, and performing offset scanning in the direction perpendicular to the weld to accommodate the assembly error of the workpiece.

[0015] Preferably, in step S2, the intelligent decision-making system matches the global point cloud data of the workpiece with the preset CAD model to extract weld information, or directly extracts weld information from the point cloud features of the workpiece's three-dimensional model.

[0016] Preferably, in step S4, the welding robot collects current and voltage data in real time during the welding process using an arc tracking sensor, and corrects the welding path in real time.

[0017] Preferably, the intelligent decision-making system integrates an AI large model and a process database. The process database adopts a two-layer architecture, including a welding process database and a weld feature database. The welding process database is used to store welding parameters under different working conditions, and the weld feature database is used to store the correspondence between weld features and welding parameters.

[0018] In summary, the present invention has the following beneficial effects: 1. The first vision system performs global scanning and reverse modeling of the workpiece. The AI ​​large model built into the intelligent decision system automatically identifies the weld position and generates the welding program without the need for manual teaching and programming. This solves the problems of complex and time-consuming programming in traditional robot welding, and reduces the professional skill threshold and labor intensity of operators.

[0019] 2. A dual-vision architecture of "first vision system + second vision system" is adopted. A large-range LiDAR is installed on the gantry truss, which can move with the gantry to perform overall scanning of large workpieces up to 20 meters long and 10 meters wide, and quickly acquire global point cloud data; an adaptive laser is installed at the welding torch end to accurately scan and locate the identified weld seams. The two work together to solve the problem of visual coverage of large workpieces and ensure the accuracy of weld seam positioning.

[0020] 3. Although the second vision system is installed at the welding torch end, it adopts a working mode of scanning before welding. Welding is only carried out after the scanning is completed, which effectively avoids the influence of arc light, spatter, high temperature and dust on the vision sensor during the welding process. It solves the problem of traditional welding torch end sensors being easily damaged and failing to recognize, and greatly improves the reliability and service life of the system.

[0021] 4. The intelligent decision-making system integrates a large AI model and a process database. Based on the basic process database, it can achieve adaptive matching of process parameters through intelligent reasoning models, solving the pain point that traditional process databases cannot exhaustively cover all operating conditions. Attached Figure Description

[0022] Figure 1 A schematic diagram of the front structure of a ship intelligent vision welding system; Figure 2 This is a schematic diagram of the bottom structure of a ship intelligent vision welding system. Figure 3 This is a side view diagram of a ship intelligent vision welding system. Figure 4 A schematic diagram of the top structure of a ship intelligent vision welding system; Figure 5 A physical image of a large-range lidar for a ship intelligent vision welding system; Figure 6 This is a software scan of a ship intelligent vision welding method; Figure 7 A physical image of an industrial control computer for a ship intelligent vision welding system; Figure 8 A physical image of a point laser positioning sensor for a ship intelligent vision welding system; Figure 9 This is a scanning range diagram of a ship intelligent vision welding system.

[0023] In the diagram: 1. Gantry crane; 2. Mobile trolley; 3. Welding robot; 4. Large-range lidar; 5. Adaptive laser; 6. Intelligent decision-making system; 7. Ground track. Detailed Implementation

[0024] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings. These embodiments do not constitute a limitation on the present invention.

[0025] Please see Figures 1 to 9 The present invention provides a technical solution: a ship intelligent vision welding system and welding method, the device including a gantry 1, a mobile trolley 2, a welding robot 3, a first vision system, a second vision system and an intelligent decision-making system 6.

[0026] Please see Figures 1 to 9 This invention provides a technical solution: a ship intelligent vision welding system, including a gantry frame 1. The gantry frame 1 is a four-column, double-beam structure with a linear moving rail on top that can move along the Y-axis. The total width of the gantry frame 1 is 12.7m, the gantry rail gauge is 12m, and the gantry height is 4m. A ground rail 7 is located at the bottom of the gantry frame 1, the length of which is configured according to production line requirements. The gantry frame 1 moves along the ground rail 7 along the X-axis. The highest point of the top guardrail of the gantry frame 1 is 4m above the ground. The distance between the gantry frame 1 and the inner side of the semi-gantry crane is 0.56m. The narrowest point of the gantry frame 1's rail for maintenance is 0.98m from the inner side of the semi-gantry crane, effectively ensuring safe passage distances.

[0027] Please see Figures 1 to 9 This invention provides a technical solution: a ship intelligent vision welding system, including a movable trolley 2, which is mounted on a linear moving rail on the top of a gantry 1. The movable trolley 2 has a rotational degree of freedom along its Z-axis, with a rotation angle range of 0-360°. The movement of the movable trolley 2 is controlled by a robot servo external axis, and the repeatability of its positioning is ±0.5mm for the X, Y, and Z axes.

[0028] Please see Figures 1 to 9 The present invention provides a technical solution: a ship intelligent vision welding system, including a welding robot 3, which is installed on the end face of a movable trolley 2. The bottom of the welding robot 3 is 1.6m above the ground, which can cover most of the ground area. The movement of the gantry 1, the linear moving rail and the movable trolley 2 and the motion of the welding robot 3 are coordinated and controlled by the robot servo external axis.

[0029] Please see Figures 1 to 9This invention provides a technical solution: a ship intelligent vision welding system and welding method, including a first vision system. The first vision system includes a large-range lidar 4 installed on a gantry 1, used to perform an overall scan of the workpiece before welding, acquire point cloud data of the workpiece surface and reverse model it. The large-range lidar 4 has a high scanning frame rate, up to 504,600 measurement points / second, and can accurately measure fast-moving small and large objects, unaffected by their shape, color or surface characteristics. It has the characteristics of large depth of field and large dynamic range, and does not require external lighting or additional linear laser. Multiple devices can be installed side by side to expand the measurement range without mutual interference.

[0030] Please see Figures 1 to 9 This invention provides a technical solution: a ship intelligent vision welding system, including a second vision system. The second vision system includes an adaptive laser 5 installed at the welding torch end of a welding robot 3, used to accurately scan and position the identified weld seam. The adaptive laser 5 adopts a high-precision point laser positioning sensor, which has the characteristics of non-contact positioning, reduced collision risk, and high positioning accuracy. It does not require deceleration to contact the workpiece, has high positioning efficiency, small sensor size, does not affect welding rate, has a large range, is compatible with workpiece error range of ±100mm, and has higher end positioning accuracy.

[0031] Please see Figures 1 to 9 This invention provides a technical solution: a ship intelligent vision welding system, including an intelligent decision-making system 6. The intelligent decision-making system 6 is connected to a first vision system, a second vision system, and a welding robot 3 control system. The intelligent decision-making system 6 integrates an AI large model and a process database. The intelligent decision-making system 6 adopts an industrial control computer configuration, including a 6U CPCIE motherboard with an i7-6820HQ processor, 16GB of RAM, a functional board, an NVIDIA GTX3050TI graphics card, and a 1TB M.2 SSD. It has built-in WOS software to control different hardware such as the robot system, welding system, and vision system. It has rich hardware interfaces and can connect to all hardware, including robots, welding machines, vision systems, PLCs, etc., enabling one-click integration and configuration of hardware devices.

[0032] Please see Figures 1 to 9This invention provides a technical solution: a ship intelligent vision welding system. An intelligent decision-making system 6 receives overall point cloud data scanned by a first vision system, automatically identifies the weld position on the workpiece, and generates a welding program. In scenarios without a CAD model, a large-range lidar 4 performs a complete scan of the workpiece surface, processes the scan results using a self-developed algorithm, automatically extracts the weld position, and generates a robot trajectory program. In scenarios with a CAD model, the large-range lidar 4 performs a complete or partial scan of the workpiece surface, matches the scan results with the CAD model, automatically extracts the weld position, and generates a robot trajectory program.

[0033] Please see Figures 1 to 9 This invention provides a technical solution: a marine intelligent vision welding system. An intelligent decision-making system 6 controls a welding robot 3 to move near the target weld and drives a second vision system to precisely locate the start and end points of the weld to correct the welding trajectory. The intelligent decision-making system 6 has built-in functions such as weld editing and weld bead planning, weld positioning, weld measurement, and weld tracking. It can configure information on weld holes, arc-starting plates, and positioning plates for the start and end points; it can configure offset parameters; it can set scanning height offset, scanning posture adjustment, scanning direction setting, and scanning length parameters; it can set the welding posture and welding offset for the start and end points; and it can set bevel information.

[0034] Please see Figures 1 to 9 This invention provides a technical solution: an intelligent vision welding system for ships, in which the movement of the gantry 1, linear moving rails, and movable trolley 2 is controlled by external axes of a robot servo system, with a walking repeatability positioning accuracy of ±0.5mm for the X, Y, and Z axes. The servo system adopts a closed-loop control principle, using an encoder to provide real-time feedback of position information, comparing it with the commanded position, and using the error signal to drive the motor for adjustment, thereby achieving high-precision positioning.

[0035] Please see Figures 1 to 9 The present invention provides a technical solution: a ship intelligent vision welding system, wherein the Z-axis of the movable trolley 2 is rotatable, which is used to adjust the working posture of the welding robot 3 installed on it. During the welding process, the robot posture can be flexibly adjusted according to the weld type, such as flat fillet weld, vertical fillet weld, and wrap fillet weld, to meet the process requirements of different welds.

[0036] Please see Figures 1 to 9 This invention provides a technical solution: an intelligent vision welding system for ships, wherein the bottom of a welding robot 3 is 1.6m above the ground to cover workpieces on the ground. For workpieces with a length less than 20,000mm, a width less than 10,000mm, and a height less than 1050mm, the welding robot 3 can effectively cover the entire working area.

[0037] Please see Figures 1 to 9This invention provides a technical solution: an intelligent vision welding method for ships, comprising the following steps: Step S1: Global Scan and Modeling: Control the gantry 1 to move, driving the large-range lidar 4 to perform a global scan of the workpiece after point-fixed assembly, obtain the global point cloud data of the workpiece, and reconstruct the three-dimensional model of the workpiece through the intelligent decision system 6. For large workpieces with a length of 20,000 mm and a width of 10,000 mm, the scanning time is controlled within 3-5 minutes.

[0038] Step S2: Teach-free programming: The built-in AI large model of the intelligent decision system 6 automatically identifies the plate combination in the 3D model, extracts weld information, and automatically generates welding programs and coarse positioning data of welds according to the preset process database. The weld extraction rule is: two intersecting plate surfaces generate a weld. The system automatically calculates the normal vectors of the two plate surfaces of the weld and the normal vector of the base plate. The system automatically establishes a workpiece coordinate system. The origin of the coordinate system is selected from a feature point with high recognizability in the workpiece, such as a workpiece corner or a reference hole. Based on the workpiece coordinate system, the system automatically generates the position information of the weld endpoints and automatically determines the weld type (flat fillet weld, vertical fillet weld, or wrap fillet weld) according to the weld position. According to the preset weld sorting and grouping rules, the system automatically generates the sequence number and group number of the welds. At the same time, according to the workpiece processing method, which is usually one process for each workpiece flipping, all weld data are processed in sequence.

[0039] Step S3: Precise weld seam positioning. Based on coarse positioning data, the intelligent decision-making system 6 controls the welding robot 3 to move near the weld seam to be welded. Then, the adaptive laser 5 at the welding torch end is activated to precisely scan the arc initiation and extinguishing points of the weld seam, obtaining its actual spatial position and orientation. During the precise scanning process, the orientation of the welding robot 3 is adjusted so that the adaptive laser 5 is aligned with the test point at a preset tilt and rotation angle, and performs offset scanning in the direction perpendicular to the weld seam to accommodate assembly errors in the workpiece. For workpieces with assembly errors, the actual spatial position and orientation of the weld seam can be obtained through precise scanning, with an error tolerance range of ±100mm. For long weld seams, if a positioning plate is installed in the middle of the workpiece, the system can identify the position of the positioning plate and automatically divide the long weld seam into multiple short, straight weld seams to adapt to actual welding requirements.

[0040] Step S4: Welding trajectory correction and execution. The intelligent decision-making system 6 corrects the welding program generated in step S2 based on the actual position information obtained from precise scanning, and controls the welding robot 3 to perform welding operations along the corrected trajectory. During welding operations, the arc tracking sensor collects current and voltage data in real time to correct the welding path. The arc tracking sensor accurately collects welding process data, achieving precise tracking, and is compatible with both flux-cored and solid wires.

[0041] Subsequent operators can manually edit the weld data, including deleting unweldable welds, adding missing welds, modifying the weld sequence, and adjusting weld groupings and processes. The edited weld data can be exported as an encrypted data package or transmitted to the production management system via network.

[0042] Please see Figures 1 to 9 This invention provides a technical solution: an intelligent vision welding method for ships, in which the process database adopts a two-layer architecture of "basic process database + inference process database". The basic process database is established based on experimental data under fixed working conditions such as material grade, plate thickness, and bevel type. It uses a unification method to establish a mapping relationship between weld shape and welding parameters such as current, voltage, speed, and attitude. The inference process database is based on an AI inference model. On the basis of the basic data, it performs parameter decoupling, hierarchical classification, and weight adjustment, and autonomously generates process parameters for non-experimental working conditions, such as adaptive parameter adjustment when plate thickness changes, enabling the process database to have self-learning capabilities.

[0043] Please see Figures 1 to 9 This invention provides a technical solution: an intelligent visual welding method for ships. The intelligent decision-making system 6 has a built-in weld seam editing function, supporting single or batch editing of information such as start and end points, weld holes, arc striking plates, positioning plates, offsets, scanning parameters such as scanning height offset, scanning posture adjustment, and scanning direction setting, as well as welding parameters such as welding posture, welding offset, and bevel information. All information can be processed for individual weld seams or edited uniformly, and batch editing by selecting weld seams is also supported. The data information after weld seam editing can be imported and exported to ensure the usability of other parts of the software.

[0044] Please see Figures 1 to 9 This invention provides a technical solution: an intelligent vision welding method for ships. Based on robot welding speed calculations, assuming a weld leg height of 6mm, welding speeds of 6mm / s for horizontal fillet welds and 3mm / s for vertical fillet welds, and a workpiece consisting of 85% horizontal fillet welds, 15% vertical fillet welds, and 10% edge plate welds, with an arc rate of 50%, and an 8-hour shift: one robotic arm can weld a length of 86 meters per shift. Two gantry cranes with four robotic arms can complete a weld length of 344 meters per shift, or 688 meters per day (two shifts), meeting the needs of large-scale production.

[0045] Please see Figures 1 to 9This invention provides a technical solution: an intelligent vision welding method for ships. When the height of the upright plate is ≤900mm, the outer corner weld can be approximately 670mm from the edge; when 900 < upright plate height ≤1300mm, the outer corner weld can be approximately 1230mm from the edge; and when 500 < upright plate height ≤800mm, the outer corner weld can be approximately 1200mm from the edge. The welding robot 3 can automatically adjust its welding posture and position according to the height of the upright plate.

[0046] Working Principle: The large-range lidar 4 of the first vision system employs laser triangulation or time-of-flight measurement. A laser beam is emitted to illuminate the workpiece surface, and the sensor receives the reflected light signal. By calculating the time of flight or imaging position of the light signal, the three-dimensional coordinates of each point on the workpiece surface are obtained. The large-range lidar 4 is mounted on the gantry 1. As the gantry 1 moves along the X-axis, it can continuously scan a large area of ​​the workpiece, obtaining complete point cloud data. After receiving the point cloud data, the intelligent decision system 6 uses algorithms such as point cloud registration, filtering, and segmentation to stitch and fuse the point clouds obtained from multi-view scanning, reconstructing a three-dimensional digital model of the workpiece. This model realistically reflects the state of the workpiece after actual fixation and assembly, including the actual position, orientation, and possible assembly errors of the sheet metal.

[0047] The intelligent decision-making system 6's built-in AI model, trained using deep learning algorithms, can automatically identify the plate assembly relationships in a 3D model. When two plates intersect, their intersection line is the weld location. The AI ​​model can output the normal vectors of the two plates at the weld and the normal vector of the base plate, providing a basis for subsequent welding posture planning. Weld data generation is based on the workpiece coordinate system. The system automatically establishes a highly identifiable workpiece coordinate system, such as using the workpiece corner as the origin, where the position information of all weld endpoints is expressed. The system automatically determines the weld type (flat fillet weld, vertical fillet weld) based on the spatial posture of the weld location and performs sorting, grouping, and sequencing according to preset rules.

[0048] The second vision system, adaptive laser 5, operates based on laser triangulation. The laser emits a linear laser beam that illuminates the weld area. A sensor receives the reflected light, and an image processing algorithm extracts the laser line's position on the image. Coordinate transformation is then used to calculate the weld's absolute position in the robot's coordinate system. During precise positioning, the system controls the welding robot 3 to move near the weld based on coarse positioning data. Then, adaptive laser 5 is activated to scan the arc initiation and extinguishing points. During scanning, the robot's tilt and rotation angles can be adjusted as needed to ensure the laser illuminates the measured point at the optimal angle. Simultaneously, an offset scan is performed perpendicular to the weld to measure its actual lateral position, accommodating workpiece assembly errors.

[0049] The first vision system, a large-range LiDAR 4, is responsible for global perception, providing coarse positioning data for the weld seam. The second vision system, an adaptive laser 5, is responsible for precise local positioning, providing the actual spatial location and orientation of the weld seam. The two systems have a clear division of labor and work collaboratively, ensuring both wide-area coverage and high-precision positioning. For applications with a CAD model, matching the scanning results with the CAD model can further improve recognition accuracy. For applications without a CAD model, the weld seam can be extracted directly from point cloud data features, and autonomous programming can also be achieved.

[0050] During the welding execution phase, the system controls the movement of welding robot 3 according to the corrected precise trajectory. Simultaneously, the arc tracking sensor collects current and voltage signals in real time during the welding process. When the welding torch deviates from the weld center, changes in the arc length cause regular fluctuations in current and voltage. The control system adjusts the robot's trajectory in real time based on these feedback signals, achieving dynamic tracking and correction during the welding process.

[0051] The movement of the gantry 1, linear guide rails, and movable trolley 2 is all controlled by the robot's servo external axes. The servo system adopts a closed-loop control principle, using encoders to provide real-time position feedback, comparing it with the commanded position, and using the error signal to drive the motor for adjustment, thereby achieving high-precision positioning with a repeatability accuracy of ±0.5mm. The Z-axis of the movable trolley 2 has rotational freedom, allowing for flexible adjustment of the welding robot 3's posture during welding to meet the process requirements of different types of welds.

[0052] The above description is merely an embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A ship intelligent vision welding system, characterized in that, include: Gantry crane, mobile trolley, welding robot, first vision system, second vision system, and intelligent decision-making system; The gantry frame is a four-column double-beam structure, with a linear moving rail on the top that can move along the Y-axis. The movable trolley is mounted on a linear moving rail on the top of the gantry, and the Z-axis of the movable trolley has a rotational degree of freedom. The welding robot is mounted on a mobile trolley and is used to weld the weld seams of the workpiece to be welded. The first vision system is mounted on a gantry and is used to perform a global scan of the workpiece to be welded and generate a point cloud model of the workpiece. The second vision system is located at the welding torch end of the welding robot and is used to accurately locate the weld seam; The intelligent decision-making system is communicatively connected to the first vision system, the second vision system, and the welding robot.

2. The intelligent vision welding system for a ship as claimed in claim 1, wherein, The gantry is equipped with a ground rail at its bottom. The gantry moves along the X-axis on the ground rail. The movement of the gantry, the linear moving rail, and the movable trolley is controlled by the robot servo external axis. Its walking repeatability positioning accuracy is ±0.5mm for the X, Y, and Z axes.

3. The intelligent vision welding system for ships according to claim 1, characterized in that, The Z-axis of the movable trolley is rotatable, which is used to adjust the working posture of the corresponding welding robot.

4. The intelligent vision welding system for ships according to claim 1, characterized in that, The bottom of the welding robot is 1.6m above the ground and is used to cover the workpiece on the ground.

5. The intelligent vision welding system for ships according to claim 1, characterized in that, The first vision system includes a large-range lidar mounted on a gantry, used to perform an overall scan of the workpiece before welding, acquire point cloud data of the workpiece surface, and perform reverse modeling. The second vision system includes an adaptive laser mounted on the welding torch of the welding robot for precise scanning and positioning of the identified weld seam.

6. A ship intelligent vision welding method based on the system described in any one of claims 1 to 5, characterized in that, Includes the following steps: S1. Control the movement of the gantry frame, and perform a global scan of the workpiece after point-fixed assembly through the first vision system to obtain the global point cloud data of the workpiece, and reverse reconstruct the three-dimensional model of the workpiece through the intelligent decision system. S2. The intelligent decision-making system automatically identifies the plate combination in the three-dimensional model of the workpiece, extracts the weld information, and automatically generates the welding program and coarse positioning data of the weld according to the preset process database. S3. The intelligent decision-making system controls the welding robot to move to the vicinity of the weld seam based on the coarse positioning data of the weld seam. The second vision system accurately scans the arc starting point and arc ending point of the weld seam to obtain the actual spatial position and posture of the weld seam. S4. The intelligent decision-making system corrects the welding program generated in step S2 based on the actual spatial position and orientation of the weld, and controls the welding robot to perform welding operations along the corrected trajectory.

7. The intelligent vision welding method for ships according to claim 6, characterized in that, In step S3, the precise scanning of the arc initiation and extinguishing points of the weld seam further includes: adjusting the posture of the welding robot so that the adaptive laser is aligned with the test point at a preset tilt angle and rotation angle, and performing offset scanning in the direction perpendicular to the weld seam to accommodate the assembly errors of the workpiece.

8. The intelligent vision welding method for ships according to claim 6, characterized in that, In step S2, the intelligent decision-making system matches the global point cloud data of the workpiece with the preset CAD model to extract weld information, or directly extracts weld information from the point cloud features of the workpiece's three-dimensional model.

9. A ship intelligent vision welding method according to claim 6, characterized in that, In step S4, the welding robot collects current and voltage data in real time during the welding process using an arc tracking sensor, and corrects the welding path in real time.

10. A ship intelligent vision welding method according to claim 6, characterized in that, The intelligent decision-making system integrates a large AI model and a process database. The process database adopts a two-layer architecture, including a welding process database and a weld feature database. The welding process database is used to store welding parameters under different working conditions, and the weld feature database is used to store the correspondence between weld features and welding parameters.

Images