Multi-robot collaborative welding system and method for automobile chassis structural parts

Abstract

The invention discloses a multi-robot collaborative welding system and method for an automobile chassis structural component, and the system comprises a sensing layer, a planning layer, a control layer, an execution layer and a monitoring layer, and the sensing layer is used for collecting a welding environment, a working state and robot motion data. The welding spot distribution is optimized through the GA-GSA algorithm, and the path is planned in combination with the improved ant colony algorithm and the DWA algorithm, so that the total welding time is shortened by 20%-30%, the robot load balance degree is improved to 95% or above, and the collaboration efficiency is remarkably improved; multi-source data fusion and fractional order PID closed-loop correction are adopted, the track deviation is controlled within + / -0.05 mm and is reduced by more than 50% compared with a traditional technology, and the positioning precision is greatly improved; through pre-generation of a static mutual exclusion region and trajectory prediction in the future 50 ms, 100% collision-free operation is realized, the deadlock problem is avoided, and the reliability of dynamic collision avoidance is enhanced.

Description

Technical Field

[0001] This invention relates to the field of automotive chassis welding technology, specifically to a multi-robot collaborative welding system and method for automotive chassis structural components. Background Technology

[0002] Automotive chassis structural components are core mechanical parts that support the overall vehicle architecture, transmit power, and enable driving functions. Their design directly affects the vehicle's handling, safety, and comfort. Automotive chassis structural components (such as longitudinal beams, cross beams, and subframes) are characterized by complex three-dimensional structures, dense welds, and high welding precision requirements (within ±0.1mm). Traditional manual welding or single-robot welding suffers from low efficiency, high labor intensity, and poor quality stability. Therefore, currently, automotive chassis structural components are welded using a multi-robot collaborative method.

[0003] In current technologies, the collaborative efficiency of system welding for automotive chassis structural components is low: weld point allocation relies on manual experience and does not consider robot load balancing and path shortestization, resulting in excessively long total welding time; the motion trajectories of multiple robots lack global planning, easily leading to path intersections and excessively long waiting times; insufficient positioning accuracy: relying solely on robot body positioning without integrating multi-source sensor data such as vision and force, it is difficult to compensate for workpiece clamping deviations and welding thermal deformation, resulting in weld trajectory deviations exceeding tolerances; weak dynamic collision avoidance capability: static collision avoidance relies on pre-programming and cannot cope with dynamic interference caused by workpiece deformation or robot motion errors, posing a collision risk; unclear collision avoidance priority rules, easily leading to deadlock when multiple robots compete for the same working area; and difficulty in quality traceability: welding process data is not stored in association with workpiece serial numbers, making it difficult to trace the root cause when quality problems occur and hindering precise process optimization. Summary of the Invention

[0004] The purpose of this invention is to provide a multi-robot collaborative welding system and method for automotive chassis structural components, in order to solve the problems of low collaborative efficiency, insufficient positioning accuracy, weak dynamic collision avoidance capability, and difficulty in quality traceability mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a multi-robot collaborative welding system and method for automotive chassis structural components, comprising a perception layer, a planning layer, a control layer, an execution layer, and a monitoring layer. The perception layer is used to collect welding environment, working status, and robot motion data. The planning layer is used for robot task allocation and path planning. The control layer is used for multi-robot motion control and collaborative logic implementation. The execution layer is used for welding action execution and auxiliary operation. The monitoring layer is used for system status monitoring and data management. The sensing layer includes a multi-source data acquisition module, a welding preprocessing module, a collision detection module, a workpiece positioning verification module, and a deviation monitoring and judgment module. The planning layer includes a weld point allocation module, a trajectory encoding and binding module, a global path planning module, a local collision avoidance optimization module, a single robot path optimization module, a cooperative state definition module, and a collision avoidance priority management module. The control layer includes a fractional-order design module, a deviation correction module, a master-slave collaborative control module, a fault-tolerant control module, a communication channel management module, and a dual-network redundancy switching module. The execution layer includes a robot motion control module, an end-effector sensor integration module, a tooling positioning module, an auxiliary support module, a welding wire supply module, a protective gas control module, a welding slag cleaning module, and a trajectory fine-tuning execution module. The monitoring layer includes a visualization module, a data storage module, a quality traceability module, a fault diagnosis module, a life prediction module, a maintenance module, an access control module, and a spare parts management module.

[0006] Preferably, the multi-source data acquisition module is used to collect multi-dimensional data such as robot motion, vision, force, and displacement, providing raw data support for subsequent processing; The welding preprocessing module is used to trim, calculate midpoints, and cluster chassis welds, reducing the computational load for subsequent planning. The collision detection module is used to pre-produce static collision-free areas, predict dynamic collision risks, and avoid collisions between the robot and tooling, workpieces, and other robots. The workpiece positioning and verification module verifies the workpiece installation position before welding to ensure positioning accuracy and avoid path deviation caused by tooling displacement. The deviation monitoring and judgment module is used to integrate multi-source data to calculate the actual path deviation, determine the deviation level according to the threshold, and trigger the corresponding processing logic.

[0007] Preferably, the weld point allocation module uses the GA-GSA algorithm to evenly distribute the chassis weld points to each robot, ensuring load balance. The trajectory encoding binding module is used to bind the solder joint allocation result with the collision-free trajectory, generate a task matrix and a trajectory matrix, and avoid trajectory conflicts after allocation. The global path planning module is used to plan the global path of multiple robots with the midpoint of the weld as the node and an improved ant colony algorithm to ensure the shortest path. The local collision avoidance optimization module is used to optimize the local trajectory based on the global path using the DWA algorithm to avoid dynamic collisions; The single-robot path optimization module is used to optimize the detail order of the midpoint path of each robot using an individual differential genetic algorithm to shorten the path length. The collaborative state definition module is used to define all working states and input events of the robot, providing a foundation for Stateflow modeling; The collision avoidance priority management module is used to formulate robot collision avoidance priority rules to resolve conflicts where multiple robots compete for mutually exclusive areas simultaneously.

[0008] Preferably, the fractional-order design module is used to design the transfer function parameters of the fractional-order PID and the integer-order fitting scheme, providing a control basis for trajectory tracking; The deviation correction module is used to receive the deviation signal from the perception layer, calculate the joint correction amount through fractional-order PID, and adjust the robot trajectory in real time. The master-slave collaborative control module is used for communication synchronization between the master controller and the robot's local controller to ensure consistent timing of multiple robot actions. The fault-tolerant control module is used to handle abnormal situations such as communication interruption and robot failure, to ensure that the system does not crash and reduce downtime losses. The communication channel management module is used to manage the communication channels between different levels, allocate bandwidth, and ensure that critical data is transmitted with priority. The dual-network redundancy switching module is used to provide redundant backup for the communication link between the main controller and the key robot, avoiding single points of failure.

[0009] Preferably, the robot motion control module is used to drive the robot joint movement to achieve precise position and attitude control of the end-effector welding torch; The end sensor integration module is used to integrate force, temperature, and vision sensors at the end of the welding torch to provide real-time feedback of welding process data. The tooling positioning module is used to ensure the installation accuracy of the chassis workpiece through modular positioning and servo correction; The auxiliary support module is used to set auxiliary supports in the easily deformable areas of the chassis to prevent the workpiece from shifting due to thermal deformation during the welding process. The welding wire supply module is used to stably supply welding wire to the welding gun, realize intelligent switching between the two wire drums, and avoid welding interruption. The protective gas control module is used to adaptively adjust the protective gas flow rate according to the welding parameters to ensure the quality of the weld formation; The slag cleaning module is used to automatically clean the welding torch nozzle and the workpiece surface of the welding gun after welding is completed, so as to avoid affecting subsequent welding. The trajectory fine-tuning execution module is used to receive the end position correction amount from the control layer, fine-tune the robot trajectory, and ensure that it conforms to the theoretical path.

[0010] Preferably, the visualization module is used to visualize the system's operating status, trajectory deviation, and fault information through the HMI interface, supporting manual viewing and intervention; The data storage module is used to store data throughout the entire system process, providing data support for traceability, diagnosis, and optimization. The quality traceability module is used to associate full-process data with chassis serial numbers, generate traceability reports, and locate quality problems. The fault diagnosis module is used to diagnose system faults based on sensor data, locate the root cause of the fault, and provide handling suggestions. The lifespan prediction module is used to predict the remaining lifespan of key components and consumables of the equipment and generate maintenance reminders in advance. The maintenance module is used to generate maintenance plans based on lifespan prediction results, record maintenance execution status, and optimize maintenance strategies. The permission management module is used to set operation permissions for different roles to avoid accidental operations and ensure system security. The spare parts management module is used to manage spare parts inventory, ensure that spare parts are available during maintenance, and support QR code scanning for warranty repairs.

[0011] Preferably, the multi-source data acquisition module includes a joint encoder data acquisition unit, an end vision data acquisition unit, a force sensor data acquisition unit, and a laser displacement data acquisition unit. The joint encoder data acquisition unit reads the absolute encoder data of the KUKA robot joints, the end vision data acquisition unit captures the relative position of the welding torch and the weld in real time, and the force sensor data acquisition unit collects the contact force between the welding torch and the workpiece during the welding process, detects the distance between the workpiece surface and the tooling reference surface, and determines whether the workpiece has shifted. The welding preprocessing module includes weld classification and trimming, weld midpoint calculation, and midpoint clustering. Weld classification and trimming involves importing the fan guard from the chassis CAD model, classifying it by length, and trimming the longer weld to two ends for intersecting welds. Weld midpoint calculation involves calculating the geometric midpoint for each weld segment using CAD coordinates and outputting a list of midpoint coordinates. Midpoint clustering uses the DBSCAN density clustering algorithm and outputs the center point of each cluster and the set of midpoints it contains. The collision detection module includes static mutual exclusion region generation and dynamic collision prediction. Static mutual exclusion region generation outputs region coordinates and prohibited joint angle range through a three-dimensional model. Dynamic collision prediction calculates the minimum distance between robots and between robots and tooling, determines the safe distance, generates a collision warning signal, and sends it to the control layer through the EtherCAT protocol. The deviation monitoring and judgment module includes multi-source data fusion and deviation classification judgment. Multi-source data fusion adopts the federated Kalman filter algorithm to fuse joint encoder data, end vision data and laser displacement data, and outputs the fused actual path coordinates. Deviation classification judgment compares the fused actual coordinates with the theoretical coordinates of the planning layer and outputs the deviation level signal. The weld point allocation module includes algorithm calculation and allocation balance verification. The algorithm calculation is to build a multi-knapsack problem model in MATLAB and output the preliminary allocation result. The allocation balance verification is to calculate the number of weld points allocated to each robot, estimate the welding time, and output the final allocation result. The trajectory encoding and binding module includes task matrix generation and trajectory matrix generation.

[0012] Preferably, the fractional-order design module includes transfer function parameter tuning and filter fitting; The master-slave collaborative control module includes state synchronization and timing coordination; The communication channel management module includes EtherCAT channel and Profinet IRT channel sub-modules.

[0013] Preferably, the robot motion control module includes a seven-axis joint drive and an end-effector posture adjustment; The tooling positioning module includes an end-point positioning pin and a servo correction; The visualization module includes 3D trajectory comparison, state machine monitoring, and deviation statistics; The quality traceability module includes traceability report generation and data association index; The lifespan prediction module includes equipment lifespan prediction and consumable lifespan prediction.

[0014] A multi-robot collaborative welding method for automotive chassis structural components includes the following steps: S1. System initialization and 3D modeling construction; S2. Workpiece clamping and positioning verification; S3. Weld pretreatment and midpoint clustering; S4. Solder joint assignment and trajectory binding; S5, Multi-level Path Planning; S6. Control parameter configuration and communication debugging; S7, Multi-robot collaborative welding execution; S8. Post-welding treatment and quality inspection; S9. Quality traceability and data storage; S10, Equipment Maintenance and Plan Updates.

[0015] Compared with the prior art, the beneficial effects of the present invention are: In this invention, the GA-GSA algorithm is used to optimize weld point allocation, and the improved ant colony algorithm and DWA algorithm are combined to plan the path, reducing the total welding time by 20%-30%, improving the robot load balance to over 95%, and significantly enhancing collaborative efficiency. Multi-source data fusion and fractional-order PID closed-loop correction are employed to control trajectory deviation within ±0.05mm, reducing it by more than 50% compared to traditional technologies, and significantly improving positioning accuracy. By pre-generating static mutually exclusive regions and predicting the trajectory 50ms in the future, combined with a priority scheduling mechanism, 100% collision-free operation is achieved, avoiding deadlock problems and enhancing dynamic collision avoidance reliability. Dual-network redundant communication and a fault task reassignment mechanism achieve system availability of 99.5%. Lifetime prediction and dynamic maintenance planning reduce downtime by 30%, optimizing system fault tolerance and maintainability. Attached Figure Description

[0016] Figure 1 This is a system diagram of the multi-robot collaborative welding system and method for automotive chassis structural components of the present invention; Figure 2 This invention relates to the sensing layer system in the multi-robot collaborative welding system and method for automotive chassis structural components. Figure 3 This is the planning layer system in the multi-robot collaborative welding system and method for automotive chassis structural components of the present invention; Figure 4 This is the control layer system in the multi-robot collaborative welding system and method for automotive chassis structural components of the present invention; Figure 5 This is the system of the execution layer in the multi-robot collaborative welding system and method for automotive chassis structural components of the present invention; Figure 6 This invention relates to a monitoring layer system for a multi-robot collaborative welding system and method for automotive chassis structural components. 1. Sensing layer; 11. Multi-source data acquisition module; 12. Welding preprocessing module; 13. Collision detection module; 14. Workpiece positioning verification module; 15. Deviation monitoring and judgment module; 2. Planning Layer; 21. Weld Point Assignment Module; 22. Trajectory Encoding and Binding Module; 23. Global Path Planning Module; 24. Local Collision Avoidance Optimization Module; 25. Single Robot Path Optimization Module; 26. Cooperative State Definition Module; 27. Collision Avoidance Priority Management Module; 3. Control Layer; 31. Fractional Order Design Module; 32. Deviation Correction Module; 33. Master-Slave Cooperative Control Module; 34. Fault-Tolerant Control Module; 35. Communication Channel Management Module; 36. Dual-Network Redundancy Switching Module; 4. Execution layer; 41. Robot motion control module; 42. End sensor integration module; 43. Tooling positioning module; 44. Auxiliary support module; 45. Welding wire supply module; 46. Shielding gas control module; 47. Welding slag cleaning module; 48. Trajectory fine-tuning execution module; 5. Monitoring layer; 51. Visualization module; 52. Data storage module; 53. Quality traceability module; 54. Fault diagnosis module; 55. Life prediction module; 56. Maintenance module; 57. Access control module; 58. Spare parts management module. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] Example 1: Refer to Figure 1 - Figure 6 As shown: A multi-robot collaborative welding system and method for automotive chassis structural components, including a perception layer 1, a planning layer 2, a control layer 3, an execution layer 4, and a monitoring layer 5. The perception layer 1 is used to collect welding environment, working status, and robot motion data. The planning layer 2 is used for robot task allocation and path planning. The control layer 3 is used for multi-robot motion control and collaborative logic implementation. The execution layer 4 is used for welding action execution and auxiliary operation. The monitoring layer 5 is used for system status monitoring and data management. The perception layer 1 includes a multi-source data acquisition module 11, a welding preprocessing module 12, a collision detection module 13, a workpiece positioning verification module 14, and a deviation monitoring and judgment module 15; The multi-source data acquisition module 11 is used to collect multi-dimensional data such as robot motion, vision, force, and displacement, providing raw data support for subsequent processing. The multi-source data acquisition module 11 includes a joint encoder data acquisition unit, an end vision data acquisition unit, a force sensing data acquisition unit, and a laser displacement data acquisition unit. The joint encoder data acquisition unit reads the absolute encoder data of the KUKA robot joints and transmits it to the edge gateway through the robot SDK interface to record the real-time angle and angular velocity of the joints. The end vision data acquisition unit installs an industrial camera and an infrared fill light on the side of each robot's welding torch to capture the relative position of the welding torch and the weld seam in real time and outputs RGB images and depth data. The force sensing data acquisition unit integrates a six-dimensional force sensor at the end of the welding torch to collect the contact force between the welding torch and the workpiece during the welding process and filter high-frequency noise. The laser displacement data acquisition unit installs a laser displacement sensor on the tooling fixture to detect the distance between the workpiece surface and the tooling reference surface in real time and determine whether the workpiece has shifted. The welding preprocessing module 12 is used to trim, calculate midpoints, and cluster chassis welds to reduce the computational load of subsequent planning. The welding preprocessing module 12 includes weld classification and trimming, weld midpoint calculation, and midpoint clustering. Weld classification and trimming involves importing fan guards from the chassis CAD model, classifying them by length, and trimming longer welds into two ends for intersecting welds to avoid movement across welds. Weld midpoint calculation involves calculating the geometric midpoint for each weld segment using CAD coordinates and outputting a list of midpoint coordinates. Midpoint clustering uses the DBSCAN density clustering algorithm to output the center point of each cluster and the set of midpoints it contains. The collision detection module 13 is used to pre-produce static collision-free areas, predict dynamic collision risks, and avoid collisions between the robot and tooling, workpieces, and other robots. The collision detection module 13 includes static mutually exclusive area generation and dynamic collision prediction. Static mutually exclusive area generation involves building a 1:1 3D model of the chassis, robot, and tooling in Process Simulate, rasterizing the workspace, marking the fixed occupancy grids of tooling and workpieces, simulating all reachable postures of the robot, generating motion sweep bodies, comparing with fixed grids, filtering out "static mutually exclusive areas that the robot cannot enter", and outputting the area coordinates and prohibited joint angle range. Dynamic collision prediction involves importing the theoretical trajectory from planning layer 2 into Process Simulate, simulating the motion trajectory of multiple robots within the next 50ms, and calculating the minimum distance between robots and between the robot and tooling. If the distance is less than the safe distance, a collision warning signal is generated and sent to control layer 3 via the EtherCAT protocol. Before welding, the workpiece positioning verification module 14 verifies the workpiece installation position to ensure positioning accuracy and avoid path deviation caused by tooling displacement. It uses a Basler Blaze1015-100 structured light 3D camera to collect point cloud data of key positioning holes on the chassis and compares it with standard CAD coordinates. If the deviation is >0.1mm, the tooling servo motor is triggered to automatically correct it. After correction, the verification is repeated until the deviation is ≤0.05mm. The deviation monitoring and judgment module 15 is used to calculate the actual path deviation by fusing multi-source data, classify the deviation level according to a threshold, and trigger the corresponding processing logic. The deviation monitoring and judgment module 15 includes multi-source data fusion and deviation classification judgment. The multi-source data fusion adopts the federated Kalman filter algorithm, fusing joint encoder data, end-effector visual data, and laser displacement data, with the weight allocation as follows: visual data 0.5, joint data 0.3, and laser data 0.2. The fused actual path coordinates are output. The deviation classification judgment compares the fused actual coordinates with the theoretical coordinates of planning layer 2 to calculate the Euclidean distance deviation. The deviation is classified according to a threshold: Level 1 deviation, Level 2 deviation, and Level 3 deviation, and the deviation level signal is output.

[0019] Planning layer 2 includes a weld point allocation module 21, a trajectory encoding and binding module 22, a global path planning module 23, a local collision avoidance optimization module 24, a single robot path optimization module 25, a cooperative state definition module 26, and a collision avoidance priority management module 27; The welding point allocation module 21 uses the GA-GSA algorithm to evenly distribute the chassis welding points to each robot to ensure load balance. The welding point allocation module 21 includes algorithm calculation and allocation balance verification. The algorithm calculation is to build a multi-knapsack problem model in MATLAB, set constraints: reachability and load balance, algorithm parameters: GA selects a percentage of 40% and iterates 300 times, and GSA gravity coefficient decays by 0.9, outputting the preliminary allocation result. The allocation balance verification is to calculate the number of welding points allocated to each robot and estimate the welding time. If the maximum time difference is >4.5%, the welding point allocation is adjusted until the time difference is ≤4.5%, and the final allocation result is output. The trajectory encoding and binding module 22 is used to bind the solder joint allocation result with the collision-free trajectory, generate a task matrix and a trajectory matrix to avoid trajectory conflicts after allocation. The trajectory encoding and binding module 22 includes task matrix generation and trajectory matrix generation. By constructing the task matrix, the elements are solder joint numbers, and the default positions are filled with 0; the solder joints are sorted according to the "region clustering" principle to reduce cross-cluster movement, and a trajectory matrix is ​​constructed that corresponds one-to-one with the task matrix, with the elements being "candidate trajectory numbers". From the collision-free trajectory library pre-generated by Process Simulate, the trajectory is matched for adjacent solder joints in the task matrix, and the shortest trajectory is selected first to ensure that there are no trajectory conflicts. The bound double matrix is ​​output. The global path planning module 23 is used to plan the global path of multiple robots with the midpoint of the weld as the node and an improved ant colony algorithm to ensure the shortest path. The local collision avoidance optimization module 24 is used to optimize the local trajectory based on the global path using the DWA algorithm to avoid dynamic collisions. It sets the linear velocity window [0, 0.4 m / s] and the angular velocity window [0, 25° / s]. The evaluation function adds a "midpoint deviation weight" to prioritize the selection of trajectories with "shortest path + midpoint fit + no collision". It receives collision warning signals in real time. If a warning is triggered, the velocity window is narrowed and the local trajectory is replanned. The single-robot path optimization module 25 is used to optimize the detail order of the midpoint path of each robot using an individual differential genetic algorithm to shorten the path length. The path is encoded in decimal integers, and the genetic strength is dynamically adjusted according to the fitness: the number of crossover genes on the chromosomes with the top 30% fitness is equal to the total number of genes × 20%, and the number of crossover genes on the chromosomes with the bottom 30% fitness is equal to the total number of genes × 60%. The selection strategy is "mixed sorting of parent and child, retaining the top 50%", and the optimal path for a single robot is output after 100 iterations. The Cooperative State Definition Module 26 is used to define all working states and input events of the robot, providing a foundation for Stateflow modeling; The collision avoidance priority management module 27 is used to formulate robot collision avoidance priority rules to resolve conflicts when multiple robots compete for mutually exclusive areas.

[0020] The control layer 3 includes a fractional-order design module 31, a deviation correction module 32, a master-slave collaborative control module 33, a fault-tolerant control module 34, a communication channel management module 35, and a dual-network redundancy switching module 36. The fractional-order design module 31 is used to design the transfer function parameters and integer-order fitting schemes of fractional-order PID controllers, providing a control foundation for trajectory tracking. The fractional-order design module 31 includes transfer function parameter tuning and filter fitting. Transfer function parameter tuning is performed in the MATLAB PID Tuner toolbox, using the "chassis corner weld trajectory" as the test object and setting the controlled object as the "robot end-effector position." The basic tuning parameter is the proportional gain. Integral constant Differential time constant Adjusting the fractional order parameter through trial and error: the order of integration. Differential order To ensure that the step response overshoot is less than 5% and the settling time is less than 0.1s, the Oustaloup filter is used for filter fitting, and the frequency domain is fitted from 1 to 100 rad / s. The fractional transfer function is fitted to a 10th-order integer transfer function. The fitting model is built in Simulink, and the fitting error is verified to be less than 3%. The integer transfer function coefficients are output for easy PLC programming. Deviation correction module 32 receives deviation signals from perception layer 1, calculates joint correction amounts using fractional-order PID control, adjusts the robot trajectory in real time, and reads the deviation vector from deviation monitoring module in real time. The end-effector position correction is calculated by substituting the integer-order transfer function; the position correction is converted into the angle correction of the 7 joints by the inverse kinematics solution of the robot; the correction is sent to the robot controller via the Profinet IRT protocol, with a correction period of ≤1ms, until the deviation returns to the first-order range, d(t)≤0.05mm; The master-slave collaborative control module 33 is used for communication synchronization between the master controller and the robot's local controller to ensure consistent timing of multiple robot actions. The master-slave collaborative control module 33 includes state synchronization and timing coordination. State synchronization involves the master controller sending a "current state command" to the robot controller via the Profinet IRT protocol; the robot controller receives this command and sends back a "state confirmation signal." After verifying all robot confirmation signals, the master controller triggers the next action, ensuring that the state synchronization deviation is ≤5ms. Timing coordination is for the symmetrical weld seams of the chassis. The master controller sets a "synchronization trigger signal," which is triggered when robot 1 arrives at... When in the specified state, a trigger signal is sent to robot 2, and robot 2 enters the state synchronously. Status: If a robot's action delay is greater than 10ms, the main controller reduces the speed of other robots and waits for the delayed robot to ensure symmetrical welding synchronization. Fault-tolerant control module 34 is used to handle abnormal situations such as communication interruption and robot failure, to ensure that the system does not crash and reduce downtime losses; The communication channel management module 35 manages the communication channels between different levels, allocates bandwidth, and ensures priority transmission of critical data. The module includes EtherCAT and Profinet IRT channel sub-modules. The perception layer 1 to control layer 3 uses the EtherCAT protocol and is divided into three channels: channel 1 transmits multi-source data acquisition results, channel 2 transmits collision warning signals, and channel 3 transmits workpiece positioning data. Channel priorities are set as follows: channel 2 > channel 1 > channel 3, ensuring priority transmission of collision warnings. The control layer 3) to execution layer 4) uses the Profinet IRT protocol and is divided into two channels: channel A transmits joint angle correction values, and channel B transmits state machine instructions. An "isochronous synchronization mode" is used to ensure that the triggering deviation of each robot action is ≤5ms. The dual-network redundancy switching module 36 is used to provide redundant backup for the communication link between the main controller and the critical robot, avoiding single points of failure. Two independent Profinet links are deployed for both the main controller and robot 1 / 2, and the communication quality of the main link is monitored in real time. If the main link packet loss rate is >10⁻ 6If the delay is greater than 1ms, the system will automatically switch to the backup link within 0.1ms. After the switch, the robot will confirm the latest instructions received by the robot through "status verification" to ensure that there is no data loss during the switch.

[0021] The execution layer 4 includes a robot motion control module 41, an end effector sensor integration module 42, a tooling positioning module 43, an auxiliary support module 44, a welding wire supply module 45, a protective gas control module 46, a welding slag cleaning module 47, and a trajectory fine-tuning execution module 48. The robot motion control module 41 is used to drive the robot joint movement and realize precise position and attitude control of the end-effector welding torch. The robot motion control module 41 includes a seven-axis joint drive and an end-effector attitude adjustment. The seven-axis joint drive receives joint angle commands from the control layer 3) and drives the seven joint motors through the servo drive unit of the KUKA KRC5 controller. It adopts a three-loop control of "position-speed-torque" to ensure the joint angle control accuracy of ±0.001° and the response time of <0.05ms. The end-effector attitude adjustment adjusts the welding torch attitude according to the weld type: the angle between the welding torch and the workpiece is 45°±5° for flat fillet welding and 60°±5° for vertical fillet welding. Through the coordinated adjustment of the robot's 6th and 7th axes, the attitude adjustment accuracy is ±1° to ensure the weld formation quality. The end sensor integration module 42 is used to integrate force, temperature, and vision sensors at the end of the welding torch to provide real-time feedback of welding process data. The welding torch end integrates: an ATI Mini45 six-dimensional force sensor; an infrared temperature sensor; and a miniature vision camera. The sensor data is transmitted to the control layer 3 via the EthCAT protocol to provide data for PID regulation and fault diagnosis. The tooling positioning module 43 is used to ensure the installation accuracy of the chassis workpiece through modular positioning and servo correction. The tooling positioning module 43 includes end-positioning pins and servo correction. The modular positioning pins adopt a modular design and are equipped with special positioning pins for different vehicle models. The positioning pin accuracy is ±0.02mm. When changing vehicle models, the positioning pins are automatically replaced by a pneumatic device. The replacement time is ≤3min. After replacement, the pin position is verified by a laser displacement sensor. The servo correction is achieved by installing 4 servo motors at the bottom of the tooling, which receive correction commands from the workpiece positioning verification module 14 and drive the tooling to make fine adjustments along the X / Y / Z axes. During the correction process, the laser displacement data is read in real time, and the correction amount is controlled in a closed loop until the workpiece positioning deviation is ≤0.05mm. The auxiliary support module 44 is used to set auxiliary supports in easily deformable areas of the chassis to prevent the workpiece from shifting due to thermal deformation during welding. Six pneumatic auxiliary support units are set at the junction of the chassis crossbeam and longitudinal beam, U-shaped groove and other easily deformable areas. Before welding, the support units press against the workpiece. During welding, the support force is monitored by a pressure sensor. If the decrease is greater than 10%, the pneumatic valve automatically replenishes the pressure. After welding is completed, the support units are retracted to avoid interfering with the robot's movement. The welding wire supply module 45 is used to stably supply welding wire to the welding torch, realize intelligent switching between the two wire spools, and avoid welding interruption. It adopts the Panasonic YW-500GD wire feeder and is equipped with two wire spools. When the main wire spool is working, the standby wire spool is in standby mode. The wire spool balance sensor monitors the main wire spool balance in real time. When the balance is <50m, it automatically switches to the standby wire spool within 0.5s. During the switching, the wire feeding motor speed is compensated to ensure continuous welding wire supply and no arc interruption. The shielding gas control module 46 is used to adaptively adjust the shielding gas flow rate according to welding parameters to ensure weld formation quality. It uses an argon / carbon dioxide mixed gas, and the flow rate is adjusted by an electromagnetic flow valve. The flow rate is 10~15L / min when the welding current is <100A, 15~20L / min when the current is ≤150A, and 20~25L / min when the current is >150A. The flow rate adjustment response time is <0.1s to ensure synchronization with current changes. The slag cleaning module 47 is used to automatically clean the welding gun nozzle and the workpiece surface of the welding after welding to avoid affecting subsequent welding. A high-pressure air gun and a rotating brush are set next to the workstation. After welding, the air gun sprays air for 10 seconds according to the preset trajectory to clean the surface slag. The rotating brush cleans the residual slag in the nozzle, with a cleaning coverage rate of ≥95%. After cleaning, the cleaning effect is checked by a vision camera. If it is not qualified, it is cleaned again. The trajectory fine-tuning execution module 48 receives the end-effector position correction amount from control layer 3, fine-tunes the robot trajectory to ensure it conforms to the theoretical path, and executes the commands sent from control layer 3. , , The correction amount is converted into joint angle fine-tuning amount through inverse kinematics of the robot; the servo drive unit performs fine-tuning action, and the fine-tuning frequency is consistent with the deviation correction cycle; after fine-tuning, the actual position is confirmed by the end vision camera to ensure that the deviation is ≤0.05mm.

[0022] The monitoring layer 5 includes a visualization module 51, a data storage module 52, a quality traceability module 53, a fault diagnosis module 54, a life prediction module 55, a maintenance module 56, an access control module 57, and a spare parts management module 58. The visualization module 51 is used to visualize the system's operating status, trajectory deviation, and fault information through the HMI interface, supporting manual viewing and intervention. The visualization module 51 includes 3D trajectory comparison, state machine monitoring, and deviation statistics. The 3D trajectory comparison imports the chassis 3D model into the WinCC interface, overlaying and displaying the theoretical path, actual path, and weld midpoint in real time; when the deviation exceeds the limit, the corresponding line segment flashes, and mouse hover is supported to view the specific deviation value. The state machine monitoring displays the current status and input events of the four robots on the left side of the interface, with color-coded changes in status; it supports viewing state transition logs for easy tracing of collaborative issues; deviation statistics use real-time statistics of "first-level / second-level / third-level deviation percentage," "current maximum deviation position," and "deviation trend curve for the past 10 minutes"; a deviation heatmap is generated by weld point number to intuitively display the deviation distribution. Data storage module 52 is used to store data from the entire system process, providing data support for traceability, diagnosis, and optimization. It adopts a MySQL industrial database and stores data including: multi-source collected data; deviation data; state machine data; and fault data. The storage period is ≥3 years for critical data and ≥1 year for general data. It supports data compression to save storage space. The quality traceability module 53 is used to associate the entire process data through the chassis serial number, generate traceability reports, and locate quality problems. The quality traceability module 53 includes traceability report generation and data association index. The injection molding report is generated by inputting the chassis serial number, which automatically associates the welding parameters, deviation data, and status records of the workpiece; it generates a PDF "quality traceability report" containing data tables, trajectory comparison charts, and deviation statistical charts, and supports export and printing. The data management index establishes a three-dimensional index of "chassis serial number - weld point number - timestamp", which allows for quick querying of associated data through any dimension, facilitating the analysis of common quality problems. The fault diagnosis module 54 is used to diagnose system faults based on sensor data, locate the root cause of the fault, provide handling suggestions, construct a fuzzy fault tree, input sensor data, calculate the fault probability of each underlying event; the diagnostic accuracy is ≥96%, and the fault root cause and handling suggestions are output. The life prediction module 55 is used to predict the remaining life of key components and consumables of the equipment and generate maintenance reminders in advance. The life prediction module 55 includes equipment life prediction and consumable life prediction. The equipment life prediction is based on the equipment operation data to build a life model and generate maintenance reminders 2 weeks in advance, displaying the remaining life and the recommended replacement time. The consumable life prediction is based on the consumption data to predict the life and generate replenishment reminders 1 day in advance. It supports automatic association with the spare parts management system. The maintenance module 56 is used to generate maintenance plans based on life prediction results, record maintenance execution status, optimize maintenance strategies, and automatically generate a "monthly maintenance plan" that includes maintenance items, planned time, and required spare parts; after maintenance is completed, operators enter the actual maintenance data; the system analyzes the maintenance effect and dynamically adjusts the maintenance cycle; The access control module (57) is used to set operation permissions for different roles to avoid accidental operations and ensure system security. It sets three levels of permissions: operator; technician; administrator; and uses password + card swipe dual authentication to record operation logs for easy tracing of accidental operations. The spare parts management module 58 is used to manage spare parts inventory, ensuring that spare parts are available during maintenance. It supports QR code-based warranty repair and repair reporting: a QR code is posted next to the equipment. When a fault occurs, the QR code is scanned, the fault type is selected, and the system is automatically uploaded to generate a repair order. Spare parts management: a spare parts database is established to record inventory quantity, supplier, and procurement cycle. When the spare parts inventory is less than the safety threshold, a purchase request is automatically generated. When the maintenance plan is generated, the spare parts inventory is checked, and if there are no spare parts, a reminder is given to purchase them in advance.

[0023] In this invention, the GA-GSA algorithm is used to optimize the allocation of weld points, and the improved ant colony algorithm and DWA algorithm are combined to plan the path, which reduces the total welding time by 20%-30%, improves the robot load balance to more than 95%, and significantly improves the collaborative efficiency. By employing multi-source data fusion and fractional-order PID closed-loop correction, the trajectory deviation is controlled within ±0.05mm, which is more than 50% lower than traditional technologies, significantly improving positioning accuracy. By pre-generating static mutually exclusive regions and predicting trajectories in the next 50ms, combined with a priority scheduling mechanism, 100% collision-free operation is achieved, deadlock problems are avoided, and the reliability of dynamic collision avoidance is enhanced. Through dual-network redundant communication and fault task redistribution mechanism, the system availability reaches 99.5%; life prediction and dynamic maintenance planning reduce downtime maintenance time by 30%, and optimize system fault tolerance and maintainability.

[0024] Example 2: A multi-robot collaborative welding method for automotive chassis structural components, specifically including the following steps: Step 1: System Initialization and 3D Modeling Construction: Establishing the "Perception-Planning-Control-Execution" hardware link: Connect the KUKA KR1000titan seven-axis robot, industrial camera, ATI Mini45 force sensor, and Keyence laser displacement sensor to the Siemens S7-1500F main controller via EtherCAT / ProfinetIRT protocol; After starting the system, check the status of each device through the monitoring layer HMI to ensure that the hardware is fault-free, and import 1:1 STEP format models of the car chassis, robot, and tooling fixtures into the Process Simulate software; Grid the workspace and mark the fixed occupied grids of the tooling / workpiece; simulate all reachable postures of the robot to generate motion sweep bodies, compare with the fixed grid to filter "static mutually exclusive regions", output the region coordinates and prohibited joint angle ranges, import them into the main controller database, and based on the static mutually exclusive regions, in Process Simulate... In Simulate, pre-plan the collision-free trajectories between each pair of welding points of the robot, and select 3-5 candidate trajectories / welding point pairs according to the principle of "shortest path and no interference" to form a "candidate trajectory library". Associate the welding point number with the trajectory number for subsequent task allocation and calling. Step 2, Workpiece Clamping and Positioning Verification: Based on the chassis model to be welded, replace the tooling positioning pins using a pneumatic device; activate the Keyence laser displacement sensor to verify that the positioning pin position deviation is ≤0.03mm. If the deviation exceeds the tolerance, fine-tune the tooling base until the accuracy requirements are met. Hoist the automotive chassis structural component onto the tooling and fix it using a servo electric clamping mechanism; activate the pneumatic auxiliary support unit in the easily deformable area of ​​the chassis to tighten the workpiece and monitor the support force, ensuring attenuation ≤10%; activate the Basler 3D camera to collect point cloud data of the key positioning holes of the chassis and compare it with the standard CAD coordinates; if the positioning deviation is >0.1mm, the main controller issues a correction command, and the tooling servo motor fine-tunes along the X / Y / Z axes until the deviation is ≤0.05mm, completing the workpiece positioning; Step 3: Weld Preprocessing and Midpoint Clustering: Extract 128 weld data points from the chassis CAD model and classify them by length: short welds are directly retained, and long welds are cut into equal-length segments of 20mm / segment; for intersecting welds, the longer welds are cut into two segments to avoid cross-region welding. Calculate the geometric midpoint for each weld segment and generate a "midpoint coordinate list". Use the DBSCAN algorithm to cluster the midpoints: set the cluster radius to 100mm and the minimum number of midpoints to 5, clustering the scattered midpoints into regional clusters, and outputting the center point of each cluster and the set of midpoints it contains, providing a regional basis for subsequent task allocation; Step 4: Weld Point Allocation and Trajectory Binding: Run the genetic-gravity search hybrid algorithm in MATLAB: Set constraints with the goal of "shortest total path and balanced load"; Algorithm parameters: GA percentage 40%, 300 iterations, GSA gravity coefficient decay 0.9, output preliminary allocation results, calculate the estimated welding time of the robots, if the maximum time difference > 4.5%, transfer 1-2 edge weld points of the heavily loaded robot to the lightly loaded robot; Finally, ensure that the deviation of the number of weld points of each robot is ≤ 1 and the time difference is ≤ 4.5%, output the final weld point allocation results, construct a "task matrix + trajectory matrix", the task matrix records the weld point number of each robot, and the trajectory matrix matches the collision-free trajectory number of adjacent weld points from the candidate trajectory library; the two matrices are associated through decoding logic to ensure "allocation is collision-free", and then sent to the main controller; Step 5, Multi-level Path Planning: Using the midpoint cluster generated in Step 3 as nodes, run the improved ant colony algorithm in MATLAB. After 200 iterations, output the global midpoint path, ensuring that the total path length is shortened by 12% compared to the traditional algorithm. The main controller calls the improved dynamic window method: setting the linear velocity window. Angular velocity window The evaluation function now includes a "midpoint deviation weight"; dynamic collision warnings from Process Simulate are read in real time, and if the distance is less than 200mm, the speed window is reduced to... The local trajectory is replanned to avoid dynamic interference. For the global path of each robot, an individual-difference genetic algorithm is used to optimize the order of details: decimal integer encoding, and the genetic strength is dynamically adjusted according to fitness. After 100 iterations, the optimal path for a single robot is output, and the average path length is shortened by 9% compared to the global path. Step Six: Control Parameter Configuration and Communication Debugging: Tune the trajectory tracking parameters in the MATLAB PID Tuner toolbox. Use an Oustaloup filter to fit the fractional-order transfer function to a 10th-order integer order, and output the coefficients to the main controller PLC for programming. Divide the EtherCAT / Profinet IRT communication channels: 3 channels for the perception layer-control layer EtherCAT protocol, and 2 channels for the control layer-execution layer Profinet IRT protocol. Enable dual-network redundancy for the link between the main controller and half of the robot, testing a switching time ≤0.1ms and a packet loss rate <10%. -6 Load finite state machine logic into the main controller: define robot state, input events, and output actions; set collision avoidance priorities to ensure that the cooperative logic is conflict-free; Step 7: Multi-robot collaborative welding execution: The operator issues a "welding start" command via HMI, and the main controller sends the initial state to the four robots according to the Stateflow logic. With the task matrix / trajectory matrix; after receiving the signal, the robot controller sends back an "acknowledgment signal," and the main controller verifies all acknowledgment signals before triggering... The event involved four robots simultaneously departing from the HOME point and entering... If the state synchronization deviation is ≤5ms, multi-source data acquisition is initiated: data is acquired from the joint encoder, end-effector vision camera, and force sensor; the data is fused using federated Kalman filtering to calculate the Euclidean distance deviation between the actual path and the theoretical path. If 0.05mm < d(t) ≤ 0.1mm, the main controller calculates the joint angle correction using a fractional-order PID controller and sends it to the robot via the ProfinetIRT protocol. The correction period is ≤ 1ms until the deviation returns to ≤ 0.05mm. If d(t) > 0.1mm, welding is paused and an alarm is triggered. The cause is manually investigated. ProcessSimulate predicts the robot's trajectory in real time for the next 50ms. If the detection interval is < 200mm, a signal is sent. The event is sent to the main controller; the main controller calls the collision avoidance priority table: if a high-priority robot occupies a mutual exclusion area, a low-priority robot can enter. The system will remain in a waiting state until the high-priority robot leaves the mutual exclusion area, at which point it will resume operation. Status; When welding symmetrical welds, the main controller sends a "synchronization trigger signal" to ensure that the two robots synchronously enter the same midpoint state. Step 8: Post-weld processing and quality inspection: After the welding task is completed, start the high-pressure air gun to clean the welding gun nozzle and both sides of the workpiece weld along the preset trajectory; at the same time, start the rotating brush to clean the residual welding slag in the nozzle; after cleaning, check the cleaning effect through the end vision camera. If the coverage rate is ≥95%, it is considered qualified. Otherwise, clean again. The 3D camera re-collects the chassis point cloud data and detects the key dimensions. If the deviation is ≤0.1mm, it is considered qualified; manually inspect the appearance of the weld and combine it with the welding force data recorded by the force sensor to determine whether the weld quality meets the standard; unqualified workpieces are marked and transferred to the rework process, and qualified workpieces are hoisted to the next process; Step Nine, Quality Traceability and Data Storage: The system automatically associates the serial number of the chassis to be welded, extracts the full process data of the workpiece from the MySQL database, and generates a PDF "Quality Traceability Report" containing trajectory comparison charts, deviation statistics charts, and weld inspection results. The welding data of this batch of chassis is compressed and stored, with key data marked as "long-term storage" and ordinary data marked as "short-term storage". Database redundancy backup is initiated to ensure no data loss, for use in subsequent quality analysis and process optimization. Step 10, Equipment Maintenance and Plan Update: The system automatically analyzes the equipment data from this welding process: if the joint temperature is >85℃ or the servo current exceeds the rated value by 90%, the system diagnoses the root cause of the fault using a fuzzy fault tree algorithm; based on runtime and vibration data, the system predicts the remaining lifespan of key components and displays maintenance reminders in the HMI pop-up window. The administrator generates a "monthly maintenance plan" based on the lifespan prediction results, including maintenance items, planned time, and required spare parts; the system checks the spare parts inventory, and if it is below the safety threshold, it automatically generates a purchase request; after maintenance is completed, the actual maintenance data is entered, the system analyzes the maintenance effect, and dynamically adjusts the maintenance cycle to ensure long-term stable operation of the equipment.

[0025] Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-robot collaborative welding system for automotive chassis structural components, characterized in that: It includes a perception layer (1), a planning layer (2), a control layer (3), an execution layer (4), and a monitoring layer (5). The perception layer (1) is used to collect welding environment, working status, and robot motion data. The planning layer (2) is used for robot task allocation and path planning. The control layer (3) is used for multi-robot motion control and collaborative logic implementation. The execution layer (4) is used for welding action execution and auxiliary operation. The monitoring layer (5) is used for system status monitoring and data management. The sensing layer (1) includes a multi-source data acquisition module (11), a welding preprocessing module (12), a collision detection module (13), a workpiece positioning verification module (14), and a deviation monitoring and judgment module (15). The planning layer (2) includes a weld point allocation module (21), a trajectory coding binding module (22), a global path planning module (23), a local collision avoidance optimization module (24), a single robot path optimization module (25), a cooperative state definition module (26), and a collision avoidance priority management module (27). The control layer (3) includes a fractional-order design module (31), a deviation correction module (32), a master-slave collaborative control module (33), a fault-tolerant control module (34), a communication channel management module (35), and a dual-network redundancy switching module (36). The execution layer (4) includes a robot motion control module (41), an end sensor integration module (42), a tooling positioning module (43), an auxiliary support module (44), a welding wire supply module (45), a protective gas control module (46), a welding slag cleaning module (47), and a trajectory fine-tuning execution module (48). The monitoring layer (5) includes a visualization module (51), a data storage module (52), a quality traceability module (53), a fault diagnosis module (54), a life prediction module (55), a maintenance module (56), an access control module (57), and a spare parts management module (58).

2. The multi-robot collaborative welding system for automotive chassis structural components according to claim 1, characterized in that: The multi-source data acquisition module (11) is used to acquire multi-dimensional data such as robot motion, vision, force, and displacement; The welding pretreatment module (12) is used to trim, calculate midpoints and cluster the chassis welds; The collision detection module (13) is used to pre-produce static collision-free areas and predict dynamic collision risks; The workpiece positioning and verification module (14) verifies the workpiece installation position before welding; The deviation monitoring and judgment module (15) is used to integrate multi-source data to calculate the actual path deviation, determine the deviation level according to the threshold, and trigger the corresponding processing logic.

3. The multi-robot collaborative welding system for automotive chassis structural components according to claim 1, characterized in that: The welding point allocation module (21) uses the GA-GSA algorithm to evenly distribute the chassis welding points to each robot; The trajectory encoding binding module (22) is used to bind the solder joint allocation result with the collision-free trajectory to generate a task matrix and a trajectory matrix; The global path planning module (23) is used to plan the global path of multiple robots using the midpoint of the weld as the node and an improved ant colony algorithm. The local collision avoidance optimization module (24) is used to optimize the local trajectory based on the global path using the DWA algorithm; The single robot path optimization module (25) is used to optimize the order of details of the midpoint path of each robot using an individual differential genetic algorithm; The collaborative state definition module (26) is used to define all working states and input events of the robot; The collision avoidance priority management module (27) is used to formulate robot collision avoidance priority rules.

4. The multi-robot collaborative welding system for automotive chassis structural components according to claim 2, characterized in that: The fractional-order design module (31) is used to design the transfer function parameters of the fractional-order PID and the integer-order fitting scheme; The deviation correction module (32) is used to receive the deviation signal from the sensing layer (1) and calculate the joint correction amount through fractional-order PID; The master-slave collaborative control module (33) is used for communication synchronization between the master controller and the robot's local controller; The fault-tolerant control module (34) is used to handle abnormal situations such as communication interruption and robot failure; The communication channel management module (35) is used to manage the communication channels between different levels and allocate bandwidth; The dual-network redundancy switching module (36) is used to perform redundant backup of the communication link between the main controller and the key robot.

5. The multi-robot collaborative welding system for automotive chassis structural components according to claim 4, characterized in that: The robot motion control module (41) is used to drive the robot joint movement; The end sensor integration module (42) is used to integrate force, temperature and vision sensors at the end of the welding torch; The tooling positioning module (43) is used for modular positioning and servo correction; The auxiliary support module (44) is used to provide auxiliary support in the easily deformable areas of the chassis; The welding wire supply module (45) is used to stably supply welding wire to the welding torch; The protective gas control module (46) is used to adaptively adjust the protective gas flow rate according to the welding parameters; The slag cleaning module (47) is used to automatically clean the welding torch nozzle and the workpiece surface of the welding slag after welding is completed; The trajectory fine-tuning execution module (48) is used to receive the end position correction amount from the control layer (3) and fine-tune the robot trajectory.

6. The multi-robot collaborative welding system for automotive chassis structural components according to claim 1, characterized in that: The visualization module (51) is used to visualize the system's operating status, trajectory deviation, and fault information through the HMI interface; The data storage module (52) is used to store the entire process data of the system; The quality traceability module (53) is used to associate the entire process data through the chassis serial number and generate a traceability report; The fault diagnosis module (54) is used to diagnose system faults based on sensor data and locate the root cause of the fault; The life prediction module (55) is used to predict the remaining life of key components and consumables of the equipment. The maintenance module (56) is used to generate a maintenance plan based on the life prediction results and record the maintenance execution status; The permission management module (57) is used to set the operation permissions for different roles; The spare parts management module (58) is used to manage spare parts inventory.

7. The multi-robot collaborative welding system for automotive chassis structural components according to claim 1, characterized in that: The multi-source data acquisition module (11) includes a joint encoder data acquisition unit, an end vision data acquisition unit, a force sensing data acquisition unit, and a laser displacement data acquisition unit. The joint encoder data acquisition unit reads the absolute encoder data of the KUKA robot joint. The end vision data acquisition unit captures the relative position of the welding torch and the weld in real time. The force sensing data acquisition unit collects the contact force between the welding torch and the workpiece during the welding process, detects the distance between the workpiece surface and the tooling reference surface, and determines whether the workpiece has shifted. The welding preprocessing module (12) includes weld classification and trimming, weld midpoint calculation and midpoint clustering. Weld classification and trimming involves importing the fan guards from the chassis CAD model, classifying them by length, and trimming the longer welds into two ends for intersecting welds. Weld midpoint calculation involves calculating the geometric midpoint for each weld segment using CAD coordinates and outputting a list of midpoint coordinates. Midpoint clustering uses the DBSCAN density clustering algorithm to output the center point of each cluster and the set of midpoints it contains. The collision detection module (13) includes static mutual exclusion region generation and dynamic collision prediction. Static mutual exclusion region generation outputs the region coordinates and prohibited joint angle range through a three-dimensional model. Dynamic collision prediction is to calculate the minimum distance between robots and between robots and tooling by the planning layer (2), determine the safe distance, generate a collision warning signal, and send it to the control layer (3) through the EtherCAT protocol. The deviation monitoring and judgment module (15) includes multi-source data fusion and deviation classification judgment. Multi-source data fusion adopts the federated Kalman filter algorithm to fuse joint encoder data, end vision data and laser displacement data, and outputs the fused actual path coordinates. Deviation classification judgment compares the fused actual coordinates with the theoretical coordinates of the planning layer (2) and outputs the deviation level signal. The welding point allocation module (21) includes algorithm calculation and allocation balance verification. The algorithm calculation is to build a multi-knapsack problem model in MATLAB and output the preliminary allocation result. The allocation balance verification is to calculate the number of welding points allocated to each robot, estimate the welding time, and output the final allocation result. The trajectory encoding binding module (22) includes task matrix generation and trajectory matrix generation.

8. The multi-robot collaborative welding system for automotive chassis structural components according to claim 3, characterized in that: The fractional-order design module (31) includes transfer function parameter tuning and filter fitting; The master-slave collaborative control module (33) includes state synchronization and timing coordination; The communication channel management module (35) includes EtherCAT channel and Profinet IRT channel sub-modules.

9. The multi-robot collaborative welding system for automotive chassis structural components according to claim 8, characterized in that: The robot motion control module (41) includes a seven-axis joint drive and an end-effector posture adjustment; The tooling positioning module (43) includes an end-point positioning pin and a servo correction; The visualization module (51) includes 3D trajectory comparison, state machine monitoring, and deviation statistics; The quality traceability module (53) includes traceability report generation and data association index; The life prediction module (55) includes equipment life prediction and consumable life prediction.

10. A multi-robot collaborative welding method for automotive chassis structural components, characterized in that: The multi-robot collaborative welding system for automotive chassis structural components according to any one of claims 1-9 specifically includes the following steps: S1. System initialization and 3D modeling construction; S2. Workpiece clamping and positioning verification; S3. Weld pretreatment and midpoint clustering; S4. Solder joint assignment and trajectory binding; S5, Multi-level Path Planning; S6. Control parameter configuration and communication debugging; S7, Multi-robot collaborative welding execution; S8. Post-welding treatment and quality inspection; S9. Quality traceability and data storage; S10, Equipment Maintenance and Plan Updates.

Images