A composite pipe bilateral collaborative robot welding system

Abstract

The present application relates to the technical field of collaborative robot welding, in particular to a kind of composite pipe double-sided collaborative robot welding system, including the following modules: trajectory and transfer area planning module generates global three-dimensional weld trajectory and dynamic transfer area, multi-constraint pose optimization module locks reference transfer position, point deduction and hot-state monitoring module generates early arc point, decay arc-out point and volatility, molten pool dynamic connection execution module controls double-machine connection according to volatility.The present application updates weld trajectory and transfer area by groove feature, determines transfer position by reachable margin and singular point avoidance, forms volatility by time window arc parameter, coordinates second robot arc starting and first robot current decay when volatility is not greater than process threshold, blocks early arc starting and makes first robot arc closing when volatility is greater than process threshold, makes double-machine connection match weld shape and molten pool hot-state, reduces weld discontinuity risk in transfer area.

Description

Technical Field

[0001] This invention relates to the field of collaborative robot welding technology, and in particular to a composite pipe double-sided collaborative robot welding system. Background Technology

[0002] The field of collaborative robot welding technology mainly involves robot trajectory control, weld seam recognition, welding torch posture adjustment, and pipe clamping and positioning. Typically, an articulated robot moves the welding torch along the weld seam, and visual sensors and arc tracking sensors are used to acquire the weld seam position and welding status. Traditional composite pipe robot welding systems refer to systems that use robots or circumferential welding mechanisms to weld the circumferential weld seam of composite pipes. These systems primarily address pipe end alignment, continuous circumferential welding, trajectory setting for different pipe diameters, and weld deviation correction. Typically, a single robot performs pipe-loop welding, with a circular guide rail driving the welding torch circumferentially or clamping and rotating the pipe while fixing the welding torch to complete the welding. An arc trajectory is generated using taught coordinate points, and weld seam positioning and position compensation are achieved using line structured light visual sensors, industrial cameras, and arc tracking sensors.

[0003] Double-sided robotic welding of composite pipes relies heavily on preset trajectories and fixed handover positions to complete opposite operations. Changes in bevel morphology can cause deviations between the preset trajectory and the actual weld. Fixed handover positions are difficult to balance the reach of robot joints and the risk of singular configurations. The arc initiation and termination positions before and after handover lack coordination with trajectory changes. There is a lack of linkage between welding thermal changes and handover timing. The two machines still complete arc initiation and termination according to the predetermined timing. The thermal state of the molten pool and the handover action are prone to misalignment. Deviations in the execution rhythm of the two machines can also affect the continuity of weld formation in the handover area. Summary of the Invention

[0004] This invention aims to provide a composite pipe double-sided collaborative robot welding system, which enables the two robots to form a global three-dimensional weld trajectory and dynamic junction zone that matches the actual weld seam based on real-time bevel characteristics during the welding process. It also determines the appropriate junction position by combining the reachability of the robot joints and the risk of singular configurations. At the same time, it coordinates the early arc initiation, arc attenuation, and double-machine connection control with the arc thermal state before junction, and constrains the arc establishment state and execution synchronization state during the double-machine connection process. This achieves coordinated control between the welding trajectory, junction position, molten pool thermal state, and connection actions of the two robots.

[0005] To achieve the above objectives, the present invention provides a composite pipe double-sided collaborative robot welding system, including a controller. The controller is equipped with a trajectory and junction area planning module, a multi-constraint pose optimization module, a point-to-point simulation and thermal monitoring module, and a molten pool dynamic continuity execution module.

[0006] The trajectory and junction planning module acquires the real-time bevel features from the line structured light sensor, fits the global three-dimensional weld trajectory, and delineates a dynamic junction zone with a preset overlapping arc length at the intersection of the two machines. The multi-constraint pose optimization module generates candidate points within the dynamic junction zone, calculates the inverse kinematics solution under the welding torch pose constraints, determines the achievable margin using the minimum residual angle of each joint to the physical limit, and locks the reference junction position by combining singular point avoidance. The point deduction and thermal monitoring module uses the reference junction position as a reference to reversely determine the early arc ignition point of the second robot and forwardly determine the arc decay extinguishing point of the first robot, and collects the arc parameters of a continuous time window when the first robot reaches the early arc ignition point to determine the fluctuation rate. The molten pool dynamic continuation execution module, when the fluctuation rate is not greater than the process threshold, causes the second robot to ignite the arc and move along the global three-dimensional weld trajectory, while simultaneously causing the first robot to move towards the arc decay extinguishing point and gradually reduce the current; when the fluctuation rate is greater than the process threshold, it intercepts the early arc ignition command and causes the first robot to extinguish the arc.

[0007] Furthermore, the trajectory and junction planning module includes a feature registration submodule, a trajectory fitting submodule, and a junction division submodule. The feature registration submodule acquires real-time bevel features and synchronous robot pose, correlates them according to the sampling time, and transforms them to the robot base coordinate system to generate a bevel spatial feature point set. The trajectory fitting submodule sorts the bevel spatial feature point set according to the weld direction, removes deviation points based on the directional consistency and spacing continuity of adjacent points, and performs spline fitting on the retained points to generate a global three-dimensional weld trajectory. The junction division submodule determines the trajectory intersection position based on the global three-dimensional weld trajectory and the opposing travel directions of the two robots, accumulates the arc length on both sides from the trajectory intersection position, and determines the segment boundary when the sum of the accumulated arc lengths reaches the preset overlapping arc length, forming a dynamic junction area.

[0008] As a further refinement of the dynamic handover zone update method, after obtaining the trajectory intersection position, the trajectory point sequence on both sides of the intersection position is verified according to the opposing travel directions of the two robots. The trajectory point whose sequence points to the intersection position is determined as the handover center point. Taking the handover center point as the starting position for arc length calculation, the arc length is accumulated to both sides along the global 3D weld trajectory to half of the preset overlapping arc length, determining the first boundary point and the second boundary point. The trajectory between the two boundary points is then defined as the dynamic handover zone. When the global 3D weld trajectory is updated, the trajectory point with the smallest spatial distance to the handover center point is selected as the updated handover center point, and the first and second boundary points are updated according to the same arc length rule, outputting the boundary of the dynamic handover zone.

[0009] Furthermore, the multi-constraint pose optimization module includes a candidate generation submodule, an inverse kinematics filtering submodule, and a position optimization submodule. The candidate generation submodule generates candidate points at preset arc length intervals within the dynamic handover zone and determines the target pose of the welding torch based on the trajectory tangent and bevel normal, forming a candidate pose set. The inverse kinematics filtering submodule calculates the inverse kinematics that satisfy the welding torch pose constraints, eliminates inverse kinematics where joint angles exceed physical limits, and generates reachable margins by taking the minimum remaining angle from each joint to its corresponding physical limit. The position optimization submodule calculates the minimum singular value of the kinematic mapping matrix for the retained inverse kinematics and determines the singular safety margin. Based on the reachable margin and the singular safety margin, it determines the inverse kinematics evaluation value, uses the maximum inverse kinematics evaluation value of the same candidate point as the candidate evaluation value, and selects the candidate point with the largest candidate evaluation value as the reference handover position.

[0010] When determining the inverse kinematics evaluation value, the reachability margin and singular safety margin corresponding to each retained inverse kinematics solution are mapped to reachability level value and singular level value according to preset upper and lower limits, respectively. The reachability weight and singular weight are read from the parameter configuration table, and the reachability level value and singular level value are weighted and summed to form the inverse kinematics evaluation value. When the difference between the highest candidate evaluation value and the second highest candidate evaluation value is not greater than the preset equivalent evaluation threshold, the total joint change of the corresponding inverse kinematics solution relative to the current joint angle of the robot is further compared, and the candidate point with the smaller total joint change is selected as the reference handover position.

[0011] Furthermore, the point-of-origin and thermal monitoring module includes a point-of-origin submodule, a thermal acquisition submodule, and a state assessment submodule. The point-of-origin submodule uses the reference junction position as the starting point for arc length calculation, accumulates a preset early arc-starting arc length in the reverse direction along the welding travel direction of the second robot, and accumulates a preset attenuation arc length in the forward direction along the welding travel direction of the first robot, generating an early arc-starting point and an attenuation arc-extinguishing point. The thermal acquisition submodule uses the moment the first robot reaches the early arc-starting point as the end of the window, collects welding current and arc voltage within a continuous time window according to a uniform sampling period, and forms an arc parameter sequence corresponding to the sampling time. The state assessment submodule calculates the change ratio of the difference between adjacent sampled values ​​relative to the average value of the corresponding window based on the arc parameter sequence, and determines the fluctuation rate by weighting according to preset current and voltage weights. When the fluctuation rate is not greater than the process threshold, it outputs a stable thermal balance state; otherwise, it outputs an abnormal thermal balance state.

[0012] The preset early arc ignition length and preset attenuation arc extinguishing length are determined based on the stable arc ignition advance time, current attenuation duration, the welding speed set for the second robot and the welding speed set for the first robot in the process parameter table, and the travel distance of the robot along the global three-dimensional weld trajectory within the corresponding time. The calculated early arc ignition point and attenuation arc extinguishing point are dynamically checked for their respective transition zone ranges. When any point exceeds the dynamic transition zone, the boundary point of the dynamic transition zone in the corresponding direction is determined as the correction point.

[0013] Furthermore, the dynamic weld pool continuity execution module includes an instruction determination submodule, a continuity coordination submodule, and an arc termination control submodule. The instruction determination submodule compares the volatility with a process threshold. When the volatility is not greater than the process threshold, it generates a handover permission signal. When the volatility is greater than the process threshold, it intercepts the early arc ignition instruction and generates an abnormal arc termination signal. The continuity coordination submodule issues an early arc ignition instruction to the second robot based on the handover permission signal. After confirming that the second robot has formed a stable arc, it controls the second robot to move along the global three-dimensional weld seam trajectory and simultaneously controls the first robot to move towards the arc extinguishing point and gradually decrease the welding current, generating a hot continuity state. The arc termination control submodule keeps the second robot from igniting an arc based on the abnormal arc termination signal and controls the first robot to gradually decrease the welding current until the arc is extinguished, outputting a handover termination state.

[0014] When confirming the stable arc establishment status of the second robot, the welding current and arc voltage of the second robot are continuously collected after the early arc ignition command is issued. When the welding current is continuously higher than the preset lower limit of arc current and the arc voltage is continuously within the preset stable voltage range for a preset confirmation time, an arc establishment signal is generated. When the preset arc ignition waiting time is reached and no arc establishment signal is generated, the arc ignition process of the second robot is terminated, the handover permission signal is canceled and an arc ignition failure blocking signal is generated, and the first robot is controlled to perform gradual current arc termination and output the arc ignition failure blocking status.

[0015] As a further constraint on the synchronous continuous control of the two robots, the real-time trajectory position of the first robot, the arcing state of the second robot, and the current control cycle timestamp of the controller are obtained. When the second robot forms a stable arc, the current decay command of the first robot and the trajectory movement command of the second robot are generated in the same control cycle and written with the same execution timestamp to form a pair of handover commands. The execution confirmation timestamps of the two robots after executing the pair of handover commands are obtained and the synchronization time deviation is determined. If the synchronization time deviation is not greater than the preset synchronization deviation threshold, the pair of handover commands for the next cycle are continued to be issued. If the synchronization time deviation is greater than the preset synchronization deviation threshold, the control time reference is reset with the later execution confirmation timestamp, and the pair of handover commands for the next cycle are generated according to the real-time trajectory positions of the two robots, and the synchronization handover status is output.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, a line structured light sensor is used to fit the global three-dimensional weld trajectory in real time based on the bevel features, and a dynamic junction zone is formed at the intersection of the two machines. The junction position is determined according to the actual weld space shape, reducing the influence of the preset trajectory deviation on the continuity position. The minimum residual angle of the joint to the physical limit is used in combination with singularity avoidance to lock the reference junction position, so that the junction position takes into account both the achievable margin and kinematic safety. The arc ignition point and arc extinguishing point are deduced around the reference junction position, and the fluctuation rate is determined according to the arc parameters of the continuous time window. When stable, the second robot is started and the current of the first robot is simultaneously decayed. When abnormal, the arc ignition is intercepted and the arc is extinguished in advance, so that the continuity action of the two machines is coordinated with the hot state of the molten pool, reducing the risk of discontinuity in weld formation in the junction zone. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall module interaction of a double-sided collaborative robot welding system for composite pipes. Figure 2 A schematic diagram of the global three-dimensional weld trajectory and the spatial relationship of the dynamic junction area; Figure 3 A schematic diagram illustrating the data processing relationships for multi-constraint pose optimization; Figure 4 A schematic diagram illustrating the formation of arc parameters and volatility within a continuous time window; Figure 5 This is a schematic diagram of the dynamic connection state transition of the molten pool; Figure 6 This is a schematic diagram showing the time relationship between the two robots executing sequentially. Detailed Implementation

[0019] The specific embodiments of the present invention will be described below with reference to the accompanying drawings. The present invention provides a composite pipe double-sided collaborative robot welding system, with a controller serving as the main control body for the dual-robot connection. The first robot and the second robot respectively travel from opposite directions towards the junction area along the same global three-dimensional weld seam trajectory. The controller is equipped with a trajectory and junction area planning module, a multi-constraint pose optimization module, a point-to-point simulation and thermal monitoring module, and a molten pool dynamic connection execution module. Each module operates according to the data and control chain of "weld seam spatial state formation - junction position determination - connection point simulation - pre-junction thermal judgment - second robot arc confirmation - dual-robot synchronous connection".

[0020] In this specification, the global three-dimensional weld trajectory is a continuous spatial trajectory formed by spatial registration, outlier removal, and spline fitting of real-time bevel features; the dynamic junction zone is a trajectory segment formed near the intersection of the two robots according to a preset overlapping arc length and updated as the global three-dimensional weld trajectory changes; the reference junction position is the junction reference position determined by combining the reachable and singular states of the robot joints within the dynamic junction zone; the early arc ignition point is the trajectory position obtained by accumulating a preset early arc ignition arc length in the opposite direction of the second robot's welding travel direction, starting from the reference junction position; and the attenuation arc extinguishing point is the trajectory position obtained by accumulating a preset attenuation arc length in the first robot's welding travel direction, starting from the reference junction position.

[0021] Volatility is used to characterize the degree of variation in welding current and arc voltage between adjacent samples of the first robot within a continuous time window before the handover. The handover permission signal indicates that the current thermal state meets the conditions for the second robot to enter the arc ignition confirmation process; the arc establishment signal indicates that the welding current, arc voltage, and duration of the second robot simultaneously meet the stable arc ignition confirmation conditions; the paired handover instructions indicate that the controller generates and writes the first robot current decay instruction and the second robot trajectory movement instruction with the same execution timestamp in the same control cycle.

[0022] Example 1 Please see Figure 1-6 This embodiment provides a composite pipe double-sided collaborative robot welding system. During system operation, a line structured light sensor continuously acquires real-time bevel features of the composite pipe bevel, while the controller simultaneously acquires the synchronous robot pose corresponding to each real-time bevel feature. The trajectory and junction zone planning module forms a global three-dimensional weld trajectory and dynamic junction zone based on two types of data; the multi-constraint pose optimization module forms multiple candidate poses within the dynamic junction zone and determines the reference junction position; the point-to-point deduction and thermal monitoring module deduces the early arc ignition point and attenuation arc extinguishing point from the reference junction position, and calculates the volatility when the first robot reaches the early arc ignition point; the molten pool dynamic continuity execution module determines whether the second robot is allowed to ignite an arc based on the volatility, and after allowing arc ignition, continues to determine whether the second robot has formed a stable arc. After a stable arc is formed, the system enters dual-robot synchronous continuity control.

[0023] Please see Figure 1 and Figure 2 The trajectory and junction planning module includes a feature registration submodule, a trajectory fitting submodule, and a junction division submodule.

[0024] The feature registration submodule receives the real-time bevel features output by the line structured light sensor and the corresponding synchronous robot pose. The real-time bevel features are the spatial position features of the bevel obtained by the line structured light sensor at the current sampling position, and the synchronous robot pose is the robot end-effector pose corresponding to that sampling moment. When the line structured light sensor is installed on the robot end-effector, the controller pre-stores the fixed spatial transformation relationship between the line structured light sensor coordinate system and the robot end-effector coordinate system; during robot operation, the spatial relationship between the robot end-effector coordinate system and the robot base coordinate system is determined based on the synchronous robot pose. The feature registration submodule sequentially performs the above coordinate transformations, so that the bevel features obtained at each sampling moment are uniformly represented in the robot base coordinate system.

[0025] Each real-time ramp feature corresponds to the synchronous robot pose according to the sampling time. The same sampling time forms a set of associated data, and the continuous associated data is transformed to form a set of ramp spatial feature points. Thus, ramp features obtained under different robot poses are transformed to the same spatial reference, and subsequent trajectory fitting does not directly use the discrete positions in the sensor's own coordinate system.

[0026] The trajectory fitting submodule reads the set of spatial feature points of the bevel and sorts each spatial feature point along the extension direction of the weld. During sorting, the continuity of adjacent points in the actual extension direction of the weld is maintained, so that the sorting result forms a point sequence that extends continuously from one side of the weld to the other.

[0027] The trajectory fitting submodule further identifies deviation points based on the consistency of direction and the continuity of spacing between adjacent points. The object of the direction consistency judgment is the local direction formed by adjacent spatial feature points. When the angle between two adjacent local directions is less than the preset direction deviation boundary, the corresponding point is considered to meet the direction continuity condition. When the local direction produces an abrupt change exceeding the direction deviation boundary relative to the preceding and following consecutive point sequences, the point is recorded as a candidate point for direction anomaly.

[0028] The determination of spacing continuity is based on the spatial distance between adjacent spatial feature points. The trajectory fitting submodule uses the adjacent distance between consecutive normal points as the spacing reference for the current point series, and points that significantly deviate from the current continuous spacing range are recorded as candidate points for spacing anomalies. The directional deviation boundary and the spacing continuity range remain unchanged within the same trajectory processing cycle, and their function is limited to determining whether the sampled points still belong to a continuous weld point series. Deviation points whose directional relationship or spatial spacing is inconsistent with the preceding and following continuous point series and cannot maintain the continuous extension relationship of the weld are eliminated.

[0029] After offset point processing, the trajectory fitting submodule performs spline fitting on the retained bevel spatial feature points. The spline fitting establishes a continuous curve according to the weld direction sequence of the retained points, ensuring continuity between adjacent retained points and generating a global 3D weld trajectory. The global 3D weld trajectory is represented in the robot's base coordinate system, and the position of any point on the trajectory can be determined according to the trajectory arc length.

[0030] The directional deviation boundary and the continuous range of spacing are not limited to fixed values. When establishing the trajectory processing parameters for the corresponding welding task, the directional deviation boundary is set according to the local directional variation range of a normal continuous bevel, and the continuous range of spacing is set according to the sampling period of the line structured light sensor and the adjacent spatial sampling distance during the robot acquisition process. This setting allows the continuous curvature variation of the normal bevel itself to be preserved, while isolated deviation sampling points do not participate in subsequent spline fitting.

[0031] The handover and segmentation submodule receives the global 3D weld seam trajectory and the opposing travel directions of the first and second robots. To facilitate trajectory arc length processing, the controller saves the point sequence of the global 3D weld seam trajectory according to the welding travel direction of the first robot, while the second robot travels along the opposite point sequence of the same trajectory. The handover and segmentation submodule determines the trajectory intersection position based on the trajectory position relationship when the two robots approach each other, and verifies the trajectory point sequence near the intersection position according to the travel directions of the two robots.

[0032] When the trajectory points of the first robot on one side of the intersection point point to the intersection point, and the trajectory points of the second robot on the other side also point to the intersection point, the corresponding trajectory positions are determined as the intersection center point. Thus, the intersection center point simultaneously satisfies the spatial intersection relationship and the relationship of the two robots approaching each other.

[0033] The junction division submodule uses the junction center point as the starting position for arc length calculation and accumulates the trajectory arc length on both sides of the global three-dimensional weld trajectory. When the accumulated arc length on each side reaches half of the preset overlapping arc length, the first boundary point and the second boundary point are determined respectively, and the trajectory segment between the first boundary point and the second boundary point forms a dynamic junction zone.

[0034] The preset overlap arc length is a length measured along the global three-dimensional weld trajectory, and is not replaced by the spatial straight-line distance between the first boundary point and the second boundary point. Before the corresponding welding program is executed, the preset overlap arc length is written into the controller as a handover process configuration, so that the dynamic handover area has a trajectory range that can accommodate the simulation of the connection point.

[0035] While the line structured light sensor continues to collect real-time bevel features, the trajectory fitting submodule can generate an updated global 3D weld trajectory. After the trajectory is updated, the junction division submodule searches for the trajectory point with the smallest spatial distance from the previous junction center point on the updated trajectory, and uses this trajectory point as the updated junction center point. Then, it accumulates half of the preset overlapping arc length from the updated junction center point to both sides to regenerate the first boundary point and the second boundary point.

[0036] Therefore, the dynamic handover area is not a fixed segment in the robot's base coordinate system, but is updated synchronously with the position changes of the global 3D weld trajectory. The trajectory and handover area planning module ultimately outputs the current global 3D weld trajectory, handover center point, first boundary point, and second boundary point to subsequent modules.

[0037] Please see Figures 1 to 3 The multi-constraint pose optimization module includes a candidate generation submodule, an inverse solution filtering submodule, and a position optimization submodule.

[0038] The candidate generation submodule reads the dynamic junction area and sequentially generates candidate points along the global 3D weld trajectory between the first and second boundary points according to a preset arc length interval. The preset arc length interval uses the same length unit as the global 3D weld trajectory, and the distance between adjacent candidate points is determined by the cumulative arc length of the trajectory, without using spatial straight-line distance as a substitute.

[0039] For each candidate point, the candidate generation submodule obtains the trajectory tangent of the global 3D weld trajectory at that candidate point and the corresponding bevel normal. The trajectory tangent determines the welding direction, and the bevel normal determines the attitude reference of the welding torch relative to the local geometry of the bevel. The candidate generation submodule establishes a mutually defined local attitude reference based on the trajectory tangent and the bevel normal, and forms the target attitude of the welding torch according to the welding torch attitude constraints already set in the current welding process. The spatial position of the candidate point and the corresponding target attitude of the welding torch together form a candidate pose. After processing all candidate points, a candidate pose set is formed.

[0040] The inverse kinematics filtering submodule sequentially reads each candidate pose and calculates the inverse kinematics solution that enables the welding torch to reach that candidate pose based on the corresponding robot kinematic relationships. Each inverse kinematics solution includes the target joint angles corresponding to each joint of the robot.

[0041] For an inverse kinematic solution, the inverse kinematic solution filtering submodule reads the lower and upper physical limits allowed for each joint. If the current target joint angle is lower than the lower physical limit or higher than the upper physical limit, the inverse kinematic solution is discarded. For inverse kinematic solutions that do not exceed the physical limits, the remaining angles from the target joint angle to the lower physical limit and to the upper physical limit for each joint are determined, and the smaller of the two remaining angles is taken as the remaining angle of the joint.

[0042] Subsequently, the minimum value is taken among all the remaining joint angles of an inverse kinematic solution, and this minimum value is determined as the reachability margin of the inverse kinematic solution. Therefore, as long as any joint approaches its physical limit, the reachability margin of the inverse kinematic solution decreases accordingly; only when all joints have a large remaining motion space does the reachability margin increase accordingly.

[0043] The position optimization submodule reads the stored inverse kinematics (IK) solutions and calculates the kinematic mapping matrix corresponding to each IK solution. The kinematic mapping matrix represents the local mapping relationship between robot joint motion and welding torch end effector motion, corresponding to the current robot model and current joint position. The position optimization submodule finds the minimum singular value of this kinematic mapping matrix and uses this minimum singular value to characterize the safety level of the current IK solution from a singular state. The smaller the minimum singular value, the closer the pose is to a kinematic singular state; as the minimum singular value increases, the corresponding singular safety margin increases.

[0044] To avoid the reachability margin and singular safety margin being directly included in the weighting due to their different numerical scales, the location optimization submodule maps them to reachability level values ​​and singular safety margin values ​​according to preset upper and lower limits. For the reachability margin, the lowest reachability level value is taken when it is less than or equal to the lower reachability mapping limit, and the highest reachability level value is taken when it is greater than or equal to the upper reachability mapping limit. When it is between the upper and lower limits, it is continuously mapped according to its position between the upper and lower limits. The singular safety margin adopts the same piecewise mapping principle, but uses independent singular safety mapping lower and upper limits.

[0045] The reachable mapping upper and lower limits are established based on the allowable range of motion of the robot joints, while the singular mapping upper and lower limits are established based on the range of minimum singular values ​​of the kinematic mapping matrix of the robot within the normal non-singular working region. These two sets of upper and lower limits apply to their respective evaluation quantities, without directly comparing angular quantities with kinematic singular evaluation quantities.

[0046] The location optimization submodule reads the reachability weights and singular weights from the parameter configuration table and ensures that both weights are on the same dimensionless weight scale before use. The reachability level value is multiplied by the corresponding reachability weight, and the singular level value is multiplied by the corresponding singular weight. The two results are then added together to obtain the inverse kinematics evaluation value. Higher reachability or singularity levels result in a larger inverse kinematics evaluation value, all other things being equal.

[0047] When a candidate point has multiple inverse kinematic solutions that satisfy the constraints, the position optimization submodule compares the inverse solution evaluation values ​​of all retained inverse solutions for that candidate point and takes the maximum value as the candidate evaluation value for that candidate point. After processing all candidate points, the candidate evaluation values ​​are compared, and the candidate point with the largest candidate evaluation value is determined as the current priority benchmark handover position.

[0048] When the difference between the highest and second-highest candidate evaluation values ​​is greater than a preset equivalent evaluation threshold, the candidate point corresponding to the highest candidate evaluation value is directly used as the benchmark handover position. When the difference is not greater than the preset equivalent evaluation threshold, the position optimization submodule further reads the corresponding inverse kinematics solution and the robot's current joint angle.

[0049] For each inverse kinematic solution to be compared, the absolute change of each target joint angle relative to the current joint angle is determined, and then the absolute changes of all joints are accumulated to form the total joint change. The candidate point with the smaller total joint change is determined as the benchmark junction position. Thus, when two candidate points are within the equivalent evaluation range in terms of reachability and singularity, the smaller overall joint motion change is used as a further selection criterion.

[0050] The preset equivalent evaluation threshold and the candidate evaluation values ​​use the same evaluation scale and remain unchanged during the optimization process of a single handover position. The multi-constraint pose optimization module finally outputs the reference handover position and the corresponding global three-dimensional weld trajectory position.

[0051] Please see Figure 2 and Figure 4 The point simulation and thermal monitoring module includes a point simulation submodule, a thermal acquisition submodule, and a status assessment submodule.

[0052] The point-based projection submodule uses the reference intersection position as the starting position for calculating the trajectory arc length. For the second robot, the preset early arc length is accumulated along the opposite direction of its welding travel direction, and the accumulated endpoint is determined as the early arc starting point; for the first robot, the preset attenuation arc length is accumulated along its welding travel direction, and the accumulated endpoint is determined as the attenuation arc extinguishing point.

[0053] When the first and second robots travel towards each other strictly along the same global three-dimensional weld seam trajectory, the opposite direction of the second robot's welding travel direction is consistent with that of the first robot. Therefore, both the early arc initiation point and the attenuation arc extinguishing point are located on one side of the reference intersection position along the first robot's travel direction. The trajectory arc lengths of these points from the reference intersection position are determined by the preset early arc initiation arc length and the preset attenuation arc extinguishing arc length, respectively. The order of these two points is determined by the magnitude of the two arc lengths themselves, without pre-restricting a fixed sequential order.

[0054] The point-based simulation submodule reads the stable arc ignition advance time, the second robot's set welding speed, the current decay duration, and the first robot's set welding speed from the process parameter table. The distance the second robot travels along the trajectory within the stable arc ignition advance time at its set welding speed is used as the preset advance arc length; the distance the first robot travels along the trajectory within the current decay duration at its set welding speed is used as the preset decay arc length.

[0055] When a constant welding speed is used within a given time interval, the trajectory travel distance is determined by the set welding speed and the corresponding duration. When the controller executes at multiple consecutive speed segments within the given time interval, the trajectory travel distance is obtained and accumulated according to the running time corresponding to each speed segment. Thus, the two arc lengths always correspond to the actual commanded travel distance of the robot along the global three-dimensional weld seam trajectory within a specified time.

[0056] After the points are formed, the point prediction submodule checks whether the early arc ignition point and the attenuation arc extinguishing point are located in the current dynamic junction zone. The judgment is made by comparing the arc length position of each point along the global 3D weld trajectory with the arc length range between the first boundary point and the second boundary point. If the predicted position is within the dynamic junction zone, the original prediction result is retained; if it exceeds the dynamic junction zone along the corresponding prediction direction, the boundary point of the dynamic junction zone in that direction is used as the corrected point.

[0057] Therefore, even if the preset early arc start length or preset attenuation arc end length is greater than the effective arc length allowed by the current dynamic junction zone, it will not cause the early arc start point or attenuation arc end point to leave the dynamic junction zone. After the global three-dimensional weld trajectory and dynamic junction zone are updated, the point-to-point simulation submodule reads the updated reference junction position and the boundary of the dynamic junction zone and re-performs the above arc length simulation.

[0058] The thermal acquisition submodule continuously acquires the welding current and arc voltage during the first robot welding process according to a uniform sampling period, and forms a one-to-one corresponding sampling sequence according to the sampling time. Each sampling moment contains a welding current sample value and an arc voltage sample value, and the same sampling period is used between adjacent sampling moments.

[0059] The controller simultaneously reads the real-time trajectory position of the first robot. When the first robot reaches the pre-arc starting point along the welding direction, or moves from before the pre-arc starting point to after the pre-arc starting point within a control cycle, it is determined that the first robot has reached the pre-arc starting point, and this control moment is taken as the end of the continuous time window.

[0060] The thermal acquisition submodule extracts sampling data corresponding to a preset window duration from the previously continuously saved arc parameter sequence, ensuring that this continuous time window covers a period of continuous arc burning before the first robot reaches the pre-arc ignition point. The welding current and arc voltage sampling values ​​within the continuous time window are arranged according to the same sampling time, thus forming the arc parameter sequence.

[0061] The condition assessment submodule first calculates the window average value of the welding current and the window average value of the arc voltage within the continuous time window. Then, according to the sampling time sequence, it calculates the absolute value of the difference between each pair of adjacent welding current sampling values, and forms the current change ratio by relating the absolute value of the absolute difference to the absolute value of the welding current window average value; in the same way, it forms the voltage change ratio for each pair of adjacent arc voltage sampling values.

[0062] For all current change ratios formed within a continuous time window, the average is performed according to the sampling time sequence to obtain the current fluctuation component of the current window; the average is performed for all voltage change ratios to obtain the voltage fluctuation component of the current window. Since both the current fluctuation component and the voltage fluctuation component are formed by adjacent absolute differences relative to their respective window averages, and both are dimensionless change ratios, the current difference in amperes and the voltage difference in volts are not directly added together.

[0063] The condition assessment submodule reads the preset current weight and preset voltage weight from the process parameter table. Both weights are non-negative dimensionless values, and are normalized to their sum before weighting, ensuring the sum of the normalized weights is one. The current fluctuation component is evaluated according to the preset current weight, and the voltage fluctuation component is evaluated according to the preset voltage weight. The sum of the two weighted results yields the fluctuation rate for the current continuous time window.

[0064] The process threshold adopts the same dimensionless evaluation scale as the volatility. When establishing the process parameter table corresponding to the current welding program, under the same material, groove type, welding current, arc voltage, and welding speed configuration, the volatility during the normal and stable arc burning stage is obtained according to the same continuous time window, sampling period, and volatility calculation rules as online operation. The upper limit of the volatility that can maintain the normal and stable arc burning state is written into the process parameter table as the process threshold for the corresponding welding program. When the welding program changes, the process threshold corresponding to that welding program is used, and thresholds under different evaluation conditions are not directly shared across welding programs.

[0065] When the volatility of the current continuous time window is not greater than the process threshold, the state assessment submodule outputs the thermal balance stable state and sends the volatility and thermal balance stable state to the molten pool dynamic continuation execution module; when the volatility is greater than the process threshold, it outputs the thermal balance abnormal state and sends the corresponding volatility and thermal balance abnormal state to the molten pool dynamic continuation execution module.

[0066] Please see Figure 1 , Figure 4 and Figure 5 The dynamic continuity execution module for the molten pool includes an instruction determination submodule, a continuity coordination submodule, and an arc termination control submodule.

[0067] The instruction determination submodule receives the volatility output from the status evaluation submodule and reads the process threshold corresponding to the current welding program. When the volatility is not greater than the process threshold, a handover permission signal is generated; when the volatility is greater than the process threshold, a handover permission signal is not generated, but the current early arc start command is intercepted, and an abnormal arc termination signal is generated.

[0068] Therefore, the second robot can only enter the early arc initiation process when the volatility is not greater than the process threshold; volatility exceeding the process threshold is a clear blocking condition and will not be interpreted as an allowed arc initiation state.

[0069] After receiving the handover permission signal, the follow-up coordination submodule sends an early arc ignition command to the second robot. After the early arc ignition command is issued, the second robot begins the arc establishment process, and the controller simultaneously starts the arc ignition waiting timer and continuously reads the welding current and arc voltage of the second robot according to a uniform sampling period.

[0070] The stable arc confirmation is based on three conditions: a preset lower limit for arc current, a preset stable voltage range, and a preset confirmation duration. The preset lower limit for arc current is a current lower limit condition, and the real-time welding current of the second robot must be continuously higher than this lower limit. The preset stable voltage range consists of a stable voltage lower limit and a stable voltage upper limit, and the real-time arc voltage of the second robot must be continuously within this range. The preset confirmation duration is the time during which the above current and voltage conditions must be continuously met.

[0071] The preset lower limit of the arc current and the preset stable voltage range are determined by the arc parameters corresponding to the current welding program, and are consistent with the welding process configuration currently used by the second robot. The preset confirmation time is used to exclude situations where the conditions are met in a single or short-term transient sampling but the actual arc has not yet entered a stable state.

[0072] After the early arc ignition command is issued, each new current and voltage sample pair is evaluated. When the current is continuously higher than the preset lower limit of arc current and the voltage is continuously within the preset stable voltage range, the continuous satisfaction time is accumulated. If either condition is no longer satisfied before the preset confirmation time is reached, the current continuous satisfaction time is interrupted, and the accumulation starts again from the next sampling time when both conditions are satisfied simultaneously.

[0073] When the two conditions are met consecutively for a preset confirmation period, the continuation coordination submodule generates an arc establishment signal. The generation of the arc establishment signal indicates that the second robot has formed a stable arc that meets the continuation conditions.

[0074] Before the arc establishment signal is generated, the first robot maintains its current welding operation state and does not prematurely enter the current decay stage. After the arc establishment signal is generated, the continuation coordination submodule controls the second robot to move along the current global three-dimensional weld seam trajectory, while simultaneously causing the first robot to continue moving towards the arc decay point and enter the gradual current decay process, thereby forming a hot continuation state.

[0075] The first robot's gradual current decay reduces the welding current cycle by cycle throughout the current decay period. The current change remains a continuous progression from the current welding current towards the arc extinguishing direction, without any direct jump from the normal welding current to the arc extinguishing state between adjacent control cycles. The first robot continues to move along the global three-dimensional weld seam trajectory towards the decaying arc extinguishing point until the corresponding arc termination control is completed.

[0076] The preset arc ignition waiting time starts counting from the actual moment the early arc ignition command is issued, and its meaning differs from the preset confirmation time. The preset confirmation time is used to determine whether a continuous stable current and stable voltage are sufficient to confirm arc establishment, while the preset arc ignition waiting time is used to limit the maximum time allowed from the issuance of the arc ignition command to the time when arc establishment can be waited for.

[0077] If an arc establishment signal is generated before the preset arc ignition waiting time is reached, the arc ignition failure judgment is stopped and the system enters the hot connection state. If an arc establishment signal is not generated after the preset arc ignition waiting time is reached, the connection coordination submodule terminates the current arc ignition process of the second robot, cancels the handover permission signal, and generates an arc ignition failure blocking signal.

[0078] Upon receiving the arc ignition failure interruption signal, the arc termination control submodule controls the first robot to begin gradually changing the current for arc termination from the current welding current and outputs the arc ignition failure interruption status. At this time, the second robot will no longer continue to execute the premature arc ignition process.

[0079] For abnormal arc termination signals caused by volatility exceeding the process threshold, the arc termination control submodule keeps the second robot in an arc-free state and causes the first robot to gradually reduce the welding current from the current welding state until the arc is extinguished, and finally outputs the handover termination state.

[0080] Therefore, abnormal arc termination and arc initiation failure blocking have different triggering sources. Abnormal arc termination occurs before the second robot initiates the arc, triggered by the volatility of the continuous time window before the handover with the first robot exceeding the process threshold; arc initiation failure blocking occurs after the volatility meets the allowable conditions and the early arc initiation command has been issued, triggered by the second robot failing to form an arc establishment signal within the preset arc initiation waiting time. Both are maintained in corresponding states in the controller and their triggering conditions are not mixed.

[0081] Please see Figure 5 and Figure 6Once the second robot forms a stable electric arc, the dynamic continuity execution module of the molten pool further executes the synchronous control of the two robots.

[0082] The controller acquires the real-time trajectory position of the first robot, the arcing status of the second robot, and the timestamp of the current control cycle. When the arcing status of the second robot corresponds to the arc establishment signal, the controller generates a current decay command for the first robot and a trajectory movement command for the second robot within the same control cycle.

[0083] The first robot current decay command includes the current decay control state corresponding to the current control cycle, which is used to enable the first robot to continue moving towards the decay extinguishing point and reduce the welding current according to the current decay stage. The second robot trajectory movement command includes the movement target for the next cycle on the current global three-dimensional weld trajectory, which is used to enable the second robot to continue moving along the weld trajectory in a stable arc-burning state.

[0084] The controller writes the same execution timestamp to two instructions generated in the same control cycle, thus forming a pair of handover instructions. The same execution timestamp uses the controller's unified control time base, and both the first and second robots interpret the current pair of handover instructions according to this common time base.

[0085] After executing their respective instructions, the two robots each return an execution confirmation to the controller. The controller records the execution confirmation timestamps of the first and second robots according to the same control time base, ensuring that the two timestamps are on the same time scale. The absolute time interval between the two execution confirmation timestamps is determined as the synchronization time deviation.

[0086] The controller compares the synchronization time deviation with a preset synchronization deviation threshold. When the synchronization time deviation is not greater than the preset synchronization deviation threshold, it is considered that the actual execution difference of the current pair handover instructions is still within the allowable range, and the current control time reference is maintained. In the next control cycle, the next pair handover instructions are generated based on the real-time trajectory position of the first robot, the real-time trajectory position of the second robot, and the arcing status.

[0087] When the synchronization time deviation exceeds a preset synchronization deviation threshold, the controller selects the later timestamp between the first robot's execution confirmation timestamp and the second robot's execution confirmation timestamp, and resets the time base for subsequent paired handover control using this later timestamp. After resetting the time base, the controller does not use the trajectory target from the previous cycle when execution differences occurred. Instead, it re-acquires the real-time trajectory positions of the two robots, and then generates the first robot's current attenuation command and the second robot's trajectory movement command for the next cycle based on the current trajectory positions, writing new identical execution timestamps for both.

[0088] Therefore, the execution confirmation timestamp constitutes the feedback entry point for synchronization control. When the synchronization time deviation is within the allowable range, the system continues to operate according to the original control time reference; when the synchronization time deviation exceeds the allowable range, the side that actually completed the execution later is used as the new time reference, and the next cycle instruction is re-formed based on the actual trajectory positions already reached by the two robots, avoiding the continued formation of subsequent instructions based on the mismatched old position state.

[0089] The preset synchronization deviation threshold is a time-based parameter, determined by the maximum execution time difference allowed by the current controller and robot sequential control, and uses the same time reference as the control cycle. Its comparative meaning remains unchanged within the same sequential task. When the robot control cycle configuration changes, the corresponding synchronization deviation threshold is redefined accordingly, ensuring that the synchronization time deviation and the threshold are always compared on the same time scale.

[0090] Please continue reading. Figure 1-6 Throughout the welding process, data updates between modules are transmitted sequentially according to the current control cycle. After obtaining new real-time bevel features, the trajectory and junction planning module forms the current global three-dimensional weld trajectory and dynamic junction area; the multi-constraint pose optimization module uses the current dynamic junction area to calculate the reference junction position; the point simulation and thermal monitoring module uses the same current trajectory state to simulate the early arc ignition point and arc attenuation extinguishing point.

[0091] When the global 3D weld trajectory changes before the first robot triggers the early arc ignition judgment, the updated dynamic handover zone, reference handover position, early arc ignition point, and arc attenuation extinguishing point are sequentially re-formed. Within one control cycle, each subsequent module uses the same set of trajectories and handover data already formed in that cycle to complete the judgment, avoiding the use of different spatial references before and after the update within the same control cycle.

[0092] After the first robot reaches the current pre-arc ignition point, the thermal acquisition submodule determines the arc parameter sequence with this moment as the end of a continuous time window, and the state assessment submodule forms the volatility according to the absolute change ratio of adjacent samples. The second robot's pre-arc ignition process only begins when the volatility meets the process threshold condition; the synchronization phase of the first robot's current decay and the second robot's trajectory movement only begins after the second robot forms a stable arc; if the second robot cannot form a stable arc, the handover is terminated and the first robot's arc is extinguished.

[0093] The aforementioned operational chain ensures that spatial position conditions, robot motion conditions, arc state conditions, and time synchronization conditions participate in the sequential control in a clearly defined sequence. The dynamic handover zone addresses the need for the handover zone to adapt to changes in the actual weld seam spatial morphology; the achievable margin and singular safety margin limit the reference handover position from falling into joint motion boundaries or approaching singular states; the advance arc ignition point and attenuation arc extinguishing point are converted from corresponding time and welding speed into arc length positions along the actual trajectory; the volatility rate is used to determine whether the first robot's current continuous arc burning state meets the handover allowable conditions before the second robot ignites; stable arc confirmation prevents the dual-robot continuation from proceeding solely based on the issuance of the arc ignition command; and the execution confirmation timestamp is used to provide feedback and correction for any existing dual-robot execution time differences.

[0094] The various preset parameters in the system are processed according to their physical properties. Preset overlapping arc length, preset arc length interval, preset early arc initiation arc length, and preset arc attenuation extinguishing arc length are spatial lengths measured along the global three-dimensional weld trajectory; robot joint physical limits and reachability margins are angular quantities; stable arc initiation advance time, current attenuation duration, continuous time window, unified sampling period, preset confirmation time, preset arc initiation waiting time, and preset synchronization deviation threshold are time quantities; preset lower limit of arc current is a current quantity; preset stable voltage range is a voltage quantity; volatility, reachability level value, singularity level value, and various weights are dimensionless evaluation quantities. Before entering the same evaluation, different physical quantities are converted to the corresponding unified scale in the aforementioned manner, and the original data with different dimensions are not directly weighted and added together.

[0095] During operation, the status of each control node can be verified through the controller output. The trajectory and handover zone planning module outputs the global three-dimensional weld trajectory, the first boundary point, and the second boundary point; the multi-constraint pose optimization module outputs the reference handover position; the point-to-point simulation and thermal monitoring module outputs the early arc ignition point, the arc attenuation extinguishing point, the fluctuation rate, and the thermal balance stable state or thermal balance abnormal state; the molten pool dynamic continuity execution module outputs the thermal continuity state, handover termination state, arc ignition failure blocking state, or synchronous handover state according to the running branch.

[0096] When the output is in a hot connection state, the second robot has already generated an arc establishment signal, and the first and second robots enter paired handover control. When the output is in a handover termination state, the corresponding volatility has exceeded the process threshold and the second robot remains un-arcged. When the output is in an arc ignition failure blocking state, the corresponding second robot has not generated an arc establishment signal within the preset arc ignition waiting time. When the output is in a synchronous handover state, the corresponding controller is maintaining or correcting the dual-robot connection time relationship according to the paired handover command and execution confirmation timestamp. All of the above states can be determined by the corresponding input conditions, comparison results, and control actions.

[0097] The real-time bevel features, bevel spatial feature point set, global three-dimensional weld trajectory, dynamic handover zone, reachable margin, singular safety margin, reference handover position, early arc ignition point, attenuation arc extinguishing point, arc parameter sequence, volatility, arc establishment signal, paired handover commands, and synchronization time deviation described in this embodiment retain the above meanings throughout the text. The processing results of each module have clearly defined downstream receiving objects, and each control state is generated by defined input conditions and judgment relationships, without simply replacing the processing procedure with a function name.

[0098] The trajectory processing method, level mapping method, window calculation method, arc confirmation method, timestamp synchronization method, and input / output relationships between modules described in the above embodiments are used to illustrate the possible implementations of the present invention. Those skilled in the art, without departing from the technical solutions described in the present invention and the application documents, may use equivalent data expressions, equivalent interface acceptance, equivalent processing order adjustments, equivalent module replacements, or equivalent changes to the control carrier for the same technical object. These modifications will not alter the essential technical relationships between the aforementioned spatial planning, pose optimization, thermal judgment, and dual-robot sequential control, and should fall within the scope of this patent protection.

Claims

1. A composite pipe double-sided collaborative robot welding system, comprising a controller, wherein the controller is equipped with a trajectory and junction area planning module, a multi-constraint pose optimization module, a point-to-point simulation and thermal monitoring module, and a molten pool dynamic continuity execution module; The trajectory and junction planning module acquires the real-time bevel features of the line structured light sensor, fits the global three-dimensional weld trajectory, and delineates a dynamic junction area with a preset overlapping arc length at the intersection of the two machines. The multi-constraint pose optimization module generates candidate points in the dynamic junction area, calculates the inverse kinematics solution under the welding gun pose constraint, determines the achievable margin by the minimum residual angle of each joint to the physical limit, and locks the reference junction position by combining singular point avoidance. The point-based simulation and thermal monitoring module uses the reference point as a reference to reversely determine the arc-starting point of the second robot and forwardly determine the arc-extinguishing point of the first robot. When the first robot reaches the arc-starting point, it collects the arc parameters of a continuous time window to determine the volatility. The dynamic continuous execution module of the molten pool causes the second robot to start an arc and move along the global three-dimensional weld seam trajectory when the fluctuation rate is not greater than the process threshold, and simultaneously causes the first robot to move toward the arc decay point and gradually reduce the current. When the volatility exceeds the process threshold, the early arc initiation command is intercepted and the first robot is made to retract the arc.

2. The system according to claim 1, characterized in that, The trajectory and transition zone planning module includes: The feature registration submodule acquires the real-time bevel features and synchronous robot pose, associates and transforms them to the robot base coordinate system according to the sampling time, and generates a set of spatial feature points of the bevel. The trajectory fitting submodule acquires the set of spatial feature points of the bevel, sorts them according to the weld direction, removes deviation points based on the consistency of direction and the continuity of spacing between adjacent points, performs spline fitting on the retained points, and generates the global three-dimensional weld trajectory. The handover division submodule determines the trajectory intersection position based on the global three-dimensional weld trajectory and the direction of travel of the two robots towards each other. It accumulates the arc length on both sides with the trajectory intersection position as the center. When the sum of the accumulated arc lengths reaches the preset overlapping arc length, it determines the segment boundary and generates the dynamic handover area.

3. The system according to claim 2, characterized in that, The handover partitioning submodule determines the dynamic handover area according to the following process: The trajectory intersection position is obtained, and the trajectory point sequence on both sides of the intersection position is verified according to the direction of travel of the two robots. The trajectory point whose point sequence points to the intersection position is determined as the intersection center point. Taking the intersection center point as the starting position of the arc length, the arc length is accumulated to half of the preset overlapping arc length on both sides along the global three-dimensional weld trajectory to determine the first boundary point and the second boundary point, and the trajectory between the two boundary points is determined as the dynamic intersection area. When the global three-dimensional weld trajectory is updated, the trajectory point with the smallest spatial distance from the intersection center point is selected as the updated intersection center point, and the first boundary point and the second boundary point are updated according to the same arc length rule, and the dynamic intersection zone boundary is output.

4. The system according to claim 1, characterized in that, The multi-constraint pose optimization module includes: The candidate generation submodule generates candidate points in the dynamic junction area at preset arc length intervals, determines the target posture of the welding gun based on the trajectory tangent and bevel normal, and generates a candidate pose set. The inverse kinematics filtering submodule calculates the inverse kinematics that satisfy the welding torch pose constraint, removes inverse kinematics where the joint angle exceeds the physical limit, and takes the minimum remaining angle of each joint to the corresponding physical limit to generate the reachable margin. The location optimization submodule calculates the minimum singular value of the kinematic mapping matrix for the retained inverse kinematic solution and determines the singular safety margin. Based on the reachable margin and the singular safety margin, it determines the inverse solution evaluation value, determines the maximum inverse solution evaluation value of the same candidate point as the candidate evaluation value, and selects the candidate point with the largest candidate evaluation value as the reference handover position.

5. The system according to claim 4, characterized in that, The process for determining the inverse solution evaluation value includes: Obtain the reachability margin and the singular safety margin corresponding to each retained inverse kinematic solution, and map them to reachability level values ​​and singular level values ​​according to preset upper and lower limits respectively; Read the reachable weight and singular weight from the parameter configuration table, and sum the reachable level value and the singular level value by weight to generate the inverse solution evaluation value; When the difference between the highest candidate evaluation value and the second highest candidate evaluation value is not greater than a preset equivalent evaluation threshold, the total joint change of the corresponding inverse kinematic solution relative to the robot's current joint angle is compared, and the candidate point with the smaller total joint change is selected as the reference handover position.

6. The system according to claim 1, characterized in that, The point simulation and thermal monitoring module includes: The point-based calculation submodule takes the reference intersection position as the starting position for arc length calculation, accumulates the preset early arc length in the reverse direction along the welding travel direction of the second robot, and accumulates the preset attenuation arc length in the forward direction along the welding travel direction of the first robot, thereby generating the early arc starting point and the attenuation arc extinguishing point. The hot acquisition submodule takes the moment when the first robot arrives at the pre-arc ignition point as the end of the window, collects the welding current and arc voltage within the continuous time window according to a uniform sampling period, and generates an arc parameter sequence according to the sampling time. The state assessment submodule calculates the change ratio of the difference between adjacent sampled values ​​relative to the average value of the corresponding window based on the arc parameter sequence, and determines the volatility by weighting according to preset current weight and voltage weight. When the volatility is not greater than the process threshold, it outputs a stable thermal balance state; otherwise, it outputs an abnormal thermal balance state.

7. The system according to claim 6, characterized in that, The process of determining the preset early arc ignition arc length and the preset attenuation arc extinguishing arc length includes: Obtain the stable arc advance time, current decay duration, second robot set welding speed and first robot set welding speed from the process parameter table, and determine the preset early arc start arc length and the preset decay arc extinguishing arc length according to the travel distance of the robot along the global three-dimensional weld trajectory within the corresponding time. Verify whether the calculated early arc initiation point and the attenuation arc extinguishing point are located in the dynamic junction area. When any point exceeds the dynamic junction area, determine the boundary point of the dynamic junction area in the corresponding direction as the correction point, and output the corrected early arc initiation point and the attenuation arc extinguishing point.

8. The system according to claim 1, characterized in that, The dynamic continuation execution module for the molten pool includes: The instruction determination submodule compares the volatility with the process threshold. When the volatility is not greater than the process threshold, it generates a handover permission signal. When the volatility is greater than the process threshold, it intercepts the early arc start instruction and generates an abnormal arc stop signal. The continuity coordination submodule sends the early arc ignition command to the second robot according to the handover permission signal. After confirming that the second robot has formed a stable arc, it controls the second robot to move along the global three-dimensional weld seam trajectory and simultaneously controls the first robot to move towards the arc extinguishing point and gradually decrease the welding current to generate a hot continuity state. The arc termination control submodule keeps the second robot from igniting an arc based on the abnormal arc termination signal, and controls the first robot to gradually reduce the welding current until the arc is extinguished, and outputs the handover termination status.

9. The system according to claim 8, characterized in that, The process of confirming that the second robot forms a stable electric arc includes: After issuing the early arc ignition command, the welding current and arc voltage of the second robot are continuously collected. When the welding current is continuously higher than the preset lower limit of arc current and the arc voltage is continuously in the preset stable voltage range for a preset confirmation time, an arc establishment signal is generated. If the arc establishment signal is not generated after the preset arc ignition waiting time is reached, the arc ignition process of the second robot is terminated, the handover permission signal is revoked and an arc ignition failure blocking signal is generated, the first robot is controlled to perform gradual current arc termination, and an arc ignition failure blocking status is output.

10. The system according to claim 8, characterized in that, The synchronous control process of the molten pool dynamic continuous execution module for the first robot and the second robot includes: The system obtains the real-time trajectory position of the first robot, the arcing state of the second robot, and the current control cycle timestamp of the controller. When the second robot forms a stable arc, it generates a current decay command for the first robot and a trajectory movement command for the second robot in the same control cycle, writes the same execution timestamp, and generates a pair of handover commands. The execution confirmation timestamps of the two robots after executing the pair handover instructions are obtained and the synchronization time deviation is determined. When the synchronization time deviation is not greater than the preset synchronization deviation threshold, the pair handover instructions for the next cycle are continued to be issued. When the synchronization time deviation is greater than the preset synchronization deviation threshold, the control time reference is reset with the later execution confirmation timestamp, and the pair handover command for the next cycle is generated based on the real-time trajectory positions of the two robots, and the synchronization handover status is output.

Images