Tailor welding workbench multi-station collaborative operation system and method

Abstract

The invention relates to a tailor-welding workbench multi-station collaborative operation system and method. The system comprises a central control unit and an AGV intelligent transfer unit. The central control unit is used for obtaining a task sequence distribution scheme based on a multi-station dynamic scheduling algorithm and the task instruction when the tailor-welding workbench receives the task instruction, and distributing the corresponding task instruction to multiple stations according to the task sequence distribution scheme and a time-sensitive network protocol; the AGV intelligent transfer unit is used for determining an optimal path according to the task instruction and transferring workpieces among the stations according to the optimal path; based on a multi-station dynamic scheduling algorithm and a time-sensitive network protocol, synchronization and dynamic optimization of a multi-station task and a task instruction are realized; the optimal path is generated through AGV intelligent transfer, multi-station transfer is achieved, and efficient, stable and uninterrupted operation of the production system is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of robot welding, and in particular to a multi-station collaborative work system and method for a tailor-welding workbench. BACKGROUND

[0002] With the development of the cold chain logistics industry towards high efficiency and environmental protection, as the core transportation equipment, the quality of the van body is facing unprecedented stringent requirements. The van body is usually made of large thin-walled metal plates (such as high-strength steel and aluminum alloy) and is tailor-welded, which needs to maintain excellent structural strength and air tightness under extreme temperature difference conditions of -30℃ to 50℃. The traditional manufacturing mode uses fixed tooling fixtures and independent work units, which has been difficult to adapt to the flexible production demand brought by lightweight and customization of new energy vehicles. The current industry generally faces the problem of low system collaboration efficiency in the van body tailor-welding link. Specifically, the production units (feeding, positioning, welding, detection, and transfer) lack unified intelligent scheduling and real-time collaboration mechanism, and the beat imbalance leads to work-in-process accumulation, rigid conveying easily causes secondary damage, local abnormalities easily cause the whole line to stop production, and the system overall efficiency and robustness are poor.

[0003] Therefore, it is urgent to provide a multi-station collaborative work system and method for a tailor-welding workbench to solve the technical problem that the production units in the system of the van body tailor-welding link lack unified intelligent scheduling and real-time collaboration mechanism, resulting in poor system overall efficiency and robustness. SUMMARY

[0004] Therefore, it is urgent to provide a multi-station collaborative work system and method for a tailor-welding workbench to solve the technical problem that the production units in the system of the van body tailor-welding link lack unified intelligent scheduling and real-time collaboration mechanism, resulting in poor system overall efficiency and robustness.

[0005] To solve the above problems, in a first aspect, the present application provides a multi-station collaborative work system for a tailor-welding workbench, comprising a central control unit and an AGV intelligent transfer unit. The central control unit is configured to obtain a task sequence allocation scheme based on a multi-station dynamic scheduling algorithm and a task instruction when the tailor-welding workbench receives the task instruction, and distribute corresponding task instructions to the multi-stations according to the task sequence allocation scheme and a time-sensitive network protocol. The AGV intelligent transfer unit is configured to determine an optimal path according to the task instruction, and transfer workpieces between stations according to the optimal path.

[0006] In a possible implementation manner, the multi-stations include interconnected automatic feeding stations, assembly positioning stations, multi-robot welding stations, and online detection stations.

[0007] In a possible implementation, the central control unit is further configured to perform multi-dimensional similarity calculation on the new vehicle model in the task instruction and the vehicle models in the process knowledge base, obtain a best reference vehicle model with the maximum similarity, perform transfer learning on the best reference vehicle model to obtain initial process parameters, perform digital twinning on the initial process parameters to obtain a vehicle three-dimensional model, perform virtual debugging on the vehicle three-dimensional model, output an optimal fixture layout scheme and a robot motion trajectory, perform dynamic scheduling on the optimal fixture layout scheme and the robot motion trajectory according to the multi-station dynamic scheduling algorithm, and obtain a task sequence allocation scheme.

[0008] In a possible implementation, the central control unit further includes determining corresponding fixture parameters and positioning instructions when each station receives the corresponding task instruction, changing the type of the fixture corresponding to the fixture parameters through a pneumatic-electromagnetic composite locking mechanism, and accurately positioning the fixture through the positioning instructions; the changing type includes disassembling an old fixture, installing a new fixture, fine adjustment of the pose, and system self-checking.

[0009] In a possible implementation, the central control unit includes a molten pool monitoring subunit; the central control unit is further configured to, when welding a workpiece at the multi-robot welding station, perform collaborative path planning according to the three-dimensional point cloud data of the weld of the workpiece and the dynamically reconfigurable topology layout formed by the multiple collaborative welding robots, obtain a welding area and a non-interference motion trajectory, and control the multiple collaborative welding robots to weld the workpiece according to the welding area and the non-interference motion trajectory, detect the molten pool oscillation frequency of the welding process in real time through the molten pool monitoring subunit, perform defect early warning through the molten pool oscillation frequency, and optimize the welding parameters through a mapping model of current-voltage-wire feeding speed.

[0010] In a possible implementation, the ring guide rail of the assembly positioning station is provided with a hydraulic push rod of an elastic groove; the central control unit is further configured to, when the assembly positioning station is expected to be delayed due to fixture replacement, start an elastic buffer adjustment mechanism; the elastic buffer adjustment mechanism is to calculate buffer demand according to delay data, control the hydraulic push rods of two adjacent elastic grooves on the ring guide rail to be started synchronously, lengthen the groove body according to the buffer demand to form a temporary storage area, and perform buffer through the temporary storage.

[0011] In a possible implementation, the central control unit is further configured to freeze the task queue of the abnormal station and reassign the unfinished production task through the multi-station dynamic scheduling algorithm when the AGV intelligent transfer unit detects that the station abnormally stops, and when the abnormal station is the multi-robot welding station, adjust the welding area division of each robot by dynamically reconstructing the robot topology layout, so that the AGV intelligent transfer unit re-plans the optimal path according to the reassigned task or the re-adjusted welding area.

[0012] In a possible implementation, the central control unit is further configured to project a precision positioning grid on the workpiece surface through a dual-wavelength laser when the AGV intelligent transfer unit transports the workpiece to the assembly positioning station, collect a grid image, extract feature corner points of the grid image, establish an initial deformation model of the workpiece by comparing the deviation of the theoretical coordinates and the actual coordinates of the feature corner points, determine a displacement compensation amount according to real-time monitoring data of the distributed sensing network and the initial deformation model, and compensate the displacement of the workpiece through the displacement compensation amount; the real-time monitoring data includes material elastic modulus, strain gauge readings, and environmental temperature change data.

[0013] In a possible implementation, the central control unit is further configured to divide the surface of the workpiece into a fine grid when the assembly positioning station cannot perform real-time compensation, obtain the curvature radius data of the center point of each grid through a three-dimensional scanner, calculate the curvature radius data according to a preset compression force optimization model to obtain an optimal compression force value, adjust the compression force of multiple compression heads through the optimal compression force value, ensure that each region of the workpiece is uniformly stressed and consistent with the theoretical curved surface, predict the springback trend of the workpiece after the compression force is released based on a cold work hardening compensation algorithm, material mechanical property data, and historical deformation records to obtain a reverse compensation force, and apply the reverse compensation force to the high-strength steel plate and then perform multiple iterative optimizations through an algorithm to obtain the compensated workpiece.

[0014] In a second aspect, the present application further provides a multi-station collaborative working method of a tailor-welded blank workstation, comprising: When the tailor-welded blank workstation receives a task instruction, a task sequence allocation scheme is obtained based on a multi-station dynamic scheduling algorithm and the task instruction, and the corresponding task instruction is distributed to the multi-station according to the task sequence allocation scheme and a time-sensitive network protocol. An optimal path is determined according to the task instruction, and the workpiece is transferred between stations according to the optimal path.

[0015] The beneficial effects of the present application are: the tailor-welding workbench multi-station collaborative operation system comprises a central control unit and an AGV intelligent transfer unit; the central control unit is used for obtaining a task sequence distribution scheme based on a multi-station dynamic scheduling algorithm and a task instruction when the tailor-welding workbench receives a task instruction, and distributing corresponding task instructions to the multi-station according to the task sequence distribution scheme and a time-sensitive network protocol; the AGV intelligent transfer unit is used for determining an optimal path according to the task instruction, and transferring workpieces between stations according to the optimal path; based on the multi-station dynamic scheduling algorithm and the time-sensitive network protocol, the synchronization and dynamic optimization of multi-station tasks and task instructions are realized; the optimal path is generated by the AGV intelligent transfer, and the transfer of the multi-station is realized, thereby ensuring the efficient, stable and uninterrupted operation of the production system. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 An embodiment flow diagram of the tailor-welding workbench multi-station collaborative operation system provided by the present application is provided. Figure 2 Another embodiment structure diagram of the tailor-welding workbench multi-station collaborative operation system provided by the present application is provided. Figure 3 An embodiment flow diagram of the online quality closed-loop control provided by the present application is provided. Figure 4 An embodiment structure diagram of the tailor-welding workbench multi-station collaborative operation method provided by the present application is provided. DETAILED DESCRIPTION

[0017] The preferred embodiments of the present application will be specifically described below in combination with the drawings, wherein the drawings constitute a part of the present application, and are used to illustrate the principles of the embodiments of the present application, and are not used to limit the scope of the present application.

[0018] As shown in the drawings, Figure 1 One specific embodiment of the present application discloses a tailor-welding workbench multi-station collaborative operation system 100, comprising a central control unit 101 and an AGV intelligent transfer unit 102. The central control unit 101 is used for obtaining a task sequence distribution scheme based on a multi-station dynamic scheduling algorithm and a task instruction when the tailor-welding workbench receives a task instruction, and distributing corresponding task instructions to the multi-station according to the task sequence distribution scheme and a time-sensitive network protocol. The AGV intelligent transfer unit 102 is used for determining an optimal path according to the task instruction, and transferring workpieces between stations according to the optimal path. The whole production process and collaborative operation situation of each station can also be monitored in real time through a distributed sensing network.

[0019] When the production management system issues a task instruction package containing vehicle model specifications, production quantity, process requirements, and emergency level information, the central control unit 101 first analyzes it. The task instruction can include the cold van vehicle model that needs to be automatically spliced and welded, so that the process parameters of the vehicle model to be processed can be determined according to the task instruction. The process parameters can be obtained from a database or other data. For example, the vehicle model to be processed can be matched with the vehicle model in the database, and then the process parameters of the corresponding vehicle model in the database can be obtained. The scheduling engine calls a multi-station dynamic scheduling algorithm trained based on deep reinforcement learning (DRL). The algorithm takes the real-time state of each station in the current system (such as equipment utilization, work-in-process queue length, and estimated completion time), process constraints of the workpiece (such as mandatory process sequence), and priority identification of the task as input, and performs online optimization calculation at the millisecond level. The optimization goal is to minimize the maximum completion time of the entire production line and balance the load of each station under the premise of meeting all process constraints. The algorithm finally outputs an optimal task sequence allocation scheme, which accurately specifies the start and end time window of each workpiece at each station (loading, positioning, welding, and detection). According to the generated scheme, the central control unit 101 distributes task instructions with accurate time stamps to the controllers of each station (such as robot controllers, positioning PLCs, and detection industrial computers) through the time-sensitive network (TSN) protocol. The TSN protocol ensures that critical motion control instructions and state feedback information can be delivered within a certain, microsecond (μs) delay through clock synchronization, traffic scheduling, and frame preemption mechanisms, so that multiple distributed welding robots, positioning mechanisms, and other devices can act like a single device, avoiding the risk of motion desynchronization or collision caused by communication delay or jitter.

[0020] The AGV intelligent transfer unit 102, under the command of the central control unit, moves workpieces precisely, reliably, and adaptively between physical workstations. The specific AGV onboard controller receives specific transfer task instructions (such as "take material from workstation A and transport it to workstation B") from the central control unit 101. Based on a built-in high-precision factory map, it integrates ultra-wideband (UWB) global positioning and laser SLAM local environmental perception data to calculate the optimal path under current conditions in real time. This path planning comprehensively considers path length, traffic congestion (positions of other AGVs), static obstacles, and workstation reservation status, aiming to achieve optimal global efficiency. When the AGV reaches the target workstation, it identifies specific markers on the workstation through an integrated vision positioning system, achieving a docking accuracy of ±1mm. Subsequently, its onboard electromagnetic locking device or pneumatic gripper automatically activates, firmly and non-destructively gripping and releasing the workpiece with the docking mechanism or workpiece pallet on the workstation, ensuring that the workpiece is not displaced or damaged during the transfer process. As a mobile sensing node, the AGV integrates lightweight sensors (such as cameras and inertial measurement units, IMUs) and is an important component of the distributed sensor network. During transport, it can perform preliminary monitoring of the workpiece's appearance and feed back its own status (position, battery level, fault codes), environmental status (channel accessibility), and observed workpiece status to the central control unit 101 in real time, providing first-hand data for global monitoring and dynamic scheduling. Through unified scheduling by the central control unit, the actions of multiple AGVs are coordinated, avoiding deadlocks and collisions. For example, when a welding station is about to complete its current task, the central control unit will pre-instruct an idle AGV to move to its buffer zone to wait, achieving "zero-wait" connection between processes, thereby significantly reducing work-in-process accumulation and improving the overall line cycle balance.

[0021] Compared with the prior art, the multi-station collaborative operation system of the welding workbench provided in this embodiment includes a central control unit and an AGV intelligent transfer unit 102. The central control unit 101 is used to obtain a task sequence allocation scheme based on the multi-station dynamic scheduling algorithm and the task instruction when the welding workbench receives a task instruction, and distribute the corresponding task instruction to the multi-station according to the task sequence allocation scheme and the time-sensitive network protocol. The AGV intelligent transfer unit 102 is used to determine the optimal path according to the task instruction and transfer the workpiece between the workstations according to the optimal path. Based on the multi-station dynamic scheduling algorithm and the time-sensitive network protocol, the synchronization and dynamic optimization of multi-station tasks and task instructions are realized. By generating the optimal path through AGV intelligent transfer and realizing the transfer of multiple workstations, the efficient, stable and uninterrupted operation of the production system is ensured.

[0022] In some embodiments of the present invention, the multi-station may include interconnected automatic feeding stations, assembly and positioning stations, multi-robot welding stations, and online inspection stations.

[0023] In some embodiments of the application, the central control unit 101 is also used to perform multi-dimensional similarity calculation on the new vehicle model in the task instruction and the vehicle model in the process knowledge base, obtain the best reference vehicle model with the maximum similarity, perform transfer learning on the best reference vehicle model to obtain initial process parameters, perform digital twinning on the initial process parameters to obtain a vehicle three-dimensional model, perform virtual debugging on the vehicle three-dimensional model, output an optimal fixture layout scheme and a robot motion trajectory, perform dynamic scheduling on the optimal fixture layout scheme and the robot motion trajectory according to a multi-station dynamic scheduling algorithm, and obtain a task sequence allocation scheme.

[0024] Among them, for the problems of weak change production adaptation ability, change fixture and parameter adjustment relying on manual trial and error, and long new vehicle model import cycle. A three-in-one rapid change type system of "process knowledge base + transfer learning + digital twinning virtual debugging" is constructed: process knowledge base + vehicle welding template, when a new vehicle model is imported, the most similar template is searched according to "similarity score = plate thickness matching × 0.4 + material matching × 0.3 + structure matching × 0.3", initial parameters are generated within 15 minutes through transfer learning; combined with modular quick-change fixture (RFID identification + pneumatic-electromagnetic composite locking, 3 minutes for fixture switching) and digital twinning pre-verification, the first-piece qualification rate is improved.

[0025] The embodiment of the application realizes rapid adaptation of vehicle type switching through process knowledge base migration learning and digital twin virtual debugging. Specifically, in the initial stage of importing a new vehicle type in the task instruction, the process knowledge base immediately starts a similar vehicle type retrieval process, and finds the best reference vehicle type by calculating a multi-dimensional similarity score. The score calculation formula is: Score = a x M_thickness + b x M_material + g x M_complexity, wherein Score represents the similarity score, a, b and g are weight coefficients of the plate thickness matching degree, the material type matching degree and the structure complexity matching degree (a = 0.4, b = 0.3, g = 0.3), and M_thickness, M_material and M_complexity are the matching degrees of the corresponding dimensions. According to the similarity score, the migration learning algorithm intelligently maps the historical vehicle type process data to the initial process parameters of the new vehicle type, and the migration coefficient is linearly adjusted in the range of 0.6 to 0.9. This process ensures the reasonable initialization of the initial process parameters and greatly reduces the trial and error times. The virtual debugging system starts to work after the process parameters are determined. The digital twin platform imports the three-dimensional model of the vehicle type, and the physical engine calculates the theoretical deformation amount based on the material properties and structure characteristics. The calculation formula is: delta = (sigma_y x L^2 x k) / (E x t^2), wherein delta represents the theoretical deformation amount (unit: mm), sigma_y is the yield strength of the material (unit: MPa), L is the characteristic length (unit: m), k is the stiffness coefficient (dimensionless), E is the elastic modulus (unit: GPa), and t is the plate thickness (unit: mm). This formula quantifies the influence of material and structure parameters on deformation, providing a theoretical basis for jig layout optimization. The joint angle of each welding robot is solved through inverse kinematics, and the reachability of the end effector to all weld points is verified. Through multiple simulation iterations, the optimal jig layout scheme and robot motion trajectory are output, ensuring that potential interference and process defects are eliminated before actual production.

[0026] The embodiment of the application stores more than 200 vehicle type welding parameter templates in the process knowledge base, and maps the historical vehicle type process data to the initial parameters of the new vehicle type through the migration learning algorithm. The vehicle type switching prediction model preloads the jig parameters and the welding program in the order scheduling stage, cooperates with the virtual debugging module based on the digital twin, realizes vehicle type switching within 15 minutes, and greatly improves the first-piece qualification rate. By constructing a "process knowledge base-migration learning-digital twin virtual debugging" trinity system, combined with the modular quick-change jig, the vehicle type switching time is shortened to within 15 minutes, and the first-piece qualification rate is greatly improved, perfectly supporting the mixed-line flexible production mode of multiple varieties and small batches.

[0027] In some embodiments of the present invention, the central control unit 101 further includes determining the corresponding fixture parameters and positioning instructions after each workstation receives the corresponding task instructions, changing the fixture corresponding to the fixture parameters through a pneumatic-electromagnetic composite locking mechanism, and accurately positioning the fixture through the positioning instructions; the change includes disassembling the old fixture, installing the new fixture, fine-tuning the position and posture, and system self-testing.

[0028] This invention also allows for rapid replacement via a modular quick-change fixture assembly and a pneumatic-electromagnetic composite locking mechanism. After scanning the new vehicle model code, the RFID automatic identification system automatically retrieves the corresponding fixture parameters and positioning instructions. The entire replacement process is completed within 15 minutes, including multiple steps such as old fixture removal, new fixture installation, fine-tuning of position, and system self-testing. The pneumatic system provides the initial locking force, while the electromagnetic mechanism achieves micron-level precise positioning, ensuring a fixture repeatability accuracy of 0.02 mm.

[0029] Furthermore, the central control unit 101 also includes a molten pool monitoring subunit that monitors the molten pool oscillation frequency in real time during the welding process of the first workpiece and provides defect warnings based on the molten pool oscillation frequency. When welding is completed, the first workpiece is inspected in full dimensions through an online inspection station. If the inspection result indicates a quality deviation, the initial process parameters of the optimal reference vehicle model are fine-tuned.

[0030] In the first-piece verification optimization stage of this invention embodiment, the system executes the complete welding process for the first workpiece. During welding, a real-time molten pool monitoring system collects molten pool dynamics at a sampling rate of 1200fps, and analyzes the molten pool oscillation frequency to provide timely warnings of potential defects. After welding, the online inspection station performs full-dimensional inspection on the first workpiece. If quality deviations are found, the system automatically fine-tunes the process parameters. Through this iterative optimization mechanism, the first-piece pass rate for new vehicle models is improved, ultimately achieving the technical goal of completing vehicle model switching within 15 minutes with a satisfactory first-piece pass rate.

[0031] Furthermore, such as Figure 2 As shown, the central control unit includes a molten pool monitoring subunit 1001, a quality data storage subunit 1002, a multi-sensor fusion detection subunit 1003, and a low-temperature environment heating subunit 1004.

[0032] In some embodiments of the present application, the central control unit 101 is also used for collaborative path planning according to the weld three-dimensional point cloud data of the workpiece and the dynamically reconfigurable topology layout formed by the multiple collaborative welding robots when welding the workpiece on the multi-robot welding station, obtaining the welding area and the non-interference motion trajectory, and controlling the multiple collaborative welding robots to weld the workpiece according to the welding area and the non-interference motion trajectory, detecting the weld pool oscillation frequency of the welding process in real time through the weld pool monitoring subunit 1001, and performing defect early warning through the weld pool oscillation frequency, and optimizing the welding parameters through the mapping model of current-voltage-wire feeding speed.

[0033] The multi-robot welding station in the embodiment of the present application includes four collaborative welding robots forming a dynamically reconfigurable topology layout, and the collaborative path planning module automatically divides the welding area and generates a non-interference motion trajectory according to the weld three-dimensional point cloud data. And control the multiple collaborative welding robots to weld the workpiece according to the welding area and the non-interference motion trajectory; also through the weld pool monitoring subunit, a 1200fps multi-band high-speed camera and a laser spectrum analyzer are linked, millisecond-level early warning of pores and incomplete fusion defects is realized through weld pool oscillation frequency analysis; the welding parameter self-optimization module establishes a mapping model of current-voltage-wire feeding speed based on a deep belief network, and dynamically adjusts the parameter deviation ≤1.5%.

[0034] Further, the process of "detecting the weld pool oscillation frequency of the welding process in real time through the weld pool monitoring subunit, and performing defect early warning through the weld pool oscillation frequency" belongs to the process of online quality closed-loop control.

[0035] In view of the problems of quality detection lag, defect recognition delay and data fragmentation, the present application has the following complete real-time quality management process, which realizes online self-calibration of process parameters through multi-sensor fusion detection and digital twin comparison.

[0036] As shown in Figure 3 , the present application has the following complete real-time quality management process, which realizes online self-calibration of process parameters through multi-sensor fusion detection and digital twin comparison. Figure 3The flowchart illustrates the online quality closed-loop control process. After the online inspection station is started, the multi-sensor fusion inspection subunit 1003 begins synchronous operation. A laser rangefinder performs precise scanning of the weld reinforcement at 0.2mm intervals, achieving an acquisition accuracy of 0.01mm; an infrared thermal imager captures the thermal stress distribution in the welding area at a frame rate of 30fps, achieving a temperature resolution of 0.1°C; an ultrasonic flaw detector uses phased array technology to perform omnidirectional inspection of the weld interior, identifying pores and incomplete fusion defects with diameters greater than 0.3mm; and a magnetic memory sensor assesses residual welding stress by monitoring changes in the metal's magnetic permeability. All sensor data is synchronized in time via a TSN network to ensure the consistency of data acquisition timing. After receiving the multi-sensor data, the digital twin comparison module begins multi-dimensional deviation analysis. The system first registers the measured data with the product's digital model and then calculates the comprehensive quality score. The score calculation formula is: S = w1 × |Δh| + w2 × | T|+w3×N_defects, where S represents the overall quality score (dimensionless), w1, w2, and w3 are the weighting coefficients for weld height deviation, temperature gradient, and defect density, respectively (dynamically adjusted according to the criticality of the weld), and Δh is the difference between the measured weld height and the model height (unit: mm). T represents the temperature gradient (unit: °C / mm), and N_defects represents the number of defects detected by ultrasonic testing (dimensionless). This formula quantifies multi-dimensional quality indicators to achieve a comprehensive evaluation of weld quality, providing a basis for process correction. When the comprehensive score exceeds a preset threshold, the system automatically marks the quality anomaly and triggers the process parameter self-calibration process. The welding parameter self-optimization module, based on a mapping model established by a deep belief network, completes parameter adjustment within 50ms. This network uses welding current, arc voltage, wire feed speed, molten pool temperature rise rate, and molten pool oscillation frequency as input features. Through a deep structure of three hidden layers (256 nodes per layer), it learns the process patterns from a historical database of high-quality welds. The core mechanism of parameter adjustment adopts the gradient descent method, calculating the adjustment amount of each parameter based on the quality deviation to ensure that the welding parameter deviation is controlled within 1.5%. Simultaneously, the real-time molten pool monitoring system uses a 1200fps high-speed camera linked with a laser spectral analyzer to achieve millisecond-level early warning of porosity and lack of fusion defects through molten pool oscillation frequency analysis, further ensuring the stability of welding quality.

[0037] The quality data storage subunit 1002 is started immediately after detection is completed. The system first generates a unique product identification code based on a millisecond-level timestamp, a production line station code, and a 32-bit cryptographic random sequence through an SHA-256 encryption algorithm. Then, a Merkle-Patricia hybrid hash tree is constructed, and Keccak-256 digest values are calculated for geometric size feature point clouds, thermal cycle temperature variation curves, and ultrasonic flaw detection graphs, respectively. Finally, a root hash is generated through nested HMAC-SHA3 encryption. When the blockchain network confirms that the root hash is successfully chained, the smart contract automatically triggers a three-stage traceability protocol to realize tamper-proof storage and full-process traceability of quality data.

[0038] Embodiments of the present application construct a quality closed-loop control throughout the manufacturing process: an online detection unit integrating multi-sensor fusion is used to realize full-dimensional and real-time detection of weld macro-morphology, micro-defects, and residual stress. Detection data are used to generate process correction instructions through a digital twin comparison module to drive weld parameter self-calibration, and the quality control mode is innovated from "post-judgment of waste" to "process intervention", significantly improving product consistency and reliability.

[0039] In view of the problems of quality detection lag, defect identification delay, and data fragmentation, in some embodiments of the present application, a hydraulic push rod with an elastic groove is arranged on the ring guide rail of the assembly positioning station; the central control unit 101 is also used to immediately start elastic buffering tamper-proof storage and full-process traceability when the assembly positioning station is expected to be delayed due to clamp replacement; the elastic buffering adjustment mechanism is to calculate the buffering demand according to the delay data, control the synchronous starting of the hydraulic push rods of the adjacent two elastic grooves on the ring guide rail, and extend the groove length according to the buffering demand to form a temporary storage area, and buffering is performed through temporary storage.

[0040] Embodiments of the present application realize online self-calibration of process parameters through multi-sensor fusion detection and digital twin comparison. Specifically, after the online detection station is started, the multi-sensor fusion detection subunit 1003 starts to work synchronously. The laser range finder scans the weld reinforcement with a precision of 0.2 mm, and the collection accuracy reaches 0.01 mm; the infrared thermal imager captures the thermal stress distribution of the welding area at a frame rate of 30 fps, and the temperature resolution reaches 0.1°C; the ultrasonic flaw detector uses phased array technology to detect the inside of the weld in all directions, and can identify pores and incomplete fusion defects with a diameter greater than 0.3 mm; the magnetic memory sensor evaluates the welding residual stress by monitoring the change of the magnetic permeability of the metal. All sensor data are time-synchronized through a TSN network to ensure the time sequence consistency of data acquisition.

[0041] In some embodiments of the present application, the central control unit 101 is also used to freeze the task queue of the abnormal station when the AGV intelligent transfer unit 102 monitors that the station has an abnormal shutdown, and to redistribute the unfinished production tasks through a multi-station dynamic scheduling algorithm. When the abnormal station is a multi-robot welding station, the welding area division of each robot is adjusted through dynamic reconstruction of the robot topology layout, so that the AGV intelligent transfer unit 102 re-plans the optimal path according to the redistributed tasks or the re-adjusted welding area.

[0042] The distributed sensor network of the embodiment of the present application starts to comprehensively collect various station running state data, including the equipment idle rate of the welding station, the queuing number of the detected workpieces of the detection station, the real-time position and running state of the AGV transfer system, and the occupation of each elastic buffer groove on the ring guide rail. All data are updated synchronously through the time-sensitive network (TSN) with a period of 100 ms, and a global production situation map is formed in the central control unit to provide data support for dynamic scheduling.

[0043] The multi-station dynamic scheduling algorithm is based on real-time production situation, and takes minimizing the production cycle and maximizing the equipment utilization rate as the multi-objective optimization direction, and comprehensively considers multiple constraint conditions such as station load balancing, order urgency, and equipment health status. The input parameters of the model include the current load of each station, the process requirements of the workpieces to be processed, the expected maintenance time of the equipment, etc., and the output is the optimal task sequence distribution scheme. When the system detects that the assembly positioning station is delayed due to clamp replacement, the elastic buffer adjustment mechanism is immediately started. The calculation formula of the buffer demand is: B=(T_delay×V_avg) / C_std, wherein B represents the buffer demand (dimensionless), T_delay is the delay time (estimated fault repair time, unit: min), V_avg is the average flow rate of the production line (the number of workpieces processed per minute, unit: pieces / min), and C_std is the standard groove capacity (the number of workpieces accommodated by a single groove, dimensionless). The formula quantifies the relationship between delay impact and production capacity, dynamically adjusts the buffer resources, and prevents work-in-process accumulation. After calculation, the hydraulic push rods of the adjacent two elastic grooves on the ring guide rail are started synchronously, and the groove length is extended from the reference 1.8 m to 2.5 m, forming a temporary storage area that can accommodate 3 workpieces to be processed. This dynamic buffer mechanism effectively prevents the entire line from stopping production due to a single station abnormality. When the system detects that a station has an abnormal shutdown through real-time monitoring, the time-sensitive network transmits the fault signal to the central control unit within 50 μs. The scheduling system immediately starts the emergency response mechanism, first freezes the task queue of the fault station, and then redistributes the unfinished production tasks. For a multi-robot welding station, the system adjusts the welding area division of each robot through dynamic reconstruction of the robot topology layout; for the AGV transfer system, the optimal path is re-planned to avoid the fault area.

[0044] All adjustment instructions are implemented through the TSN network to achieve microsecond-level synchronization, ensuring that the system can still maintain stable operation under abnormal working conditions. In addition, the cooperative control bus adopts a time-division multiplexing mechanism to allocate exclusive communication time slots, and the transmission delay of key motion instructions does not exceed 35 microseconds. The CRC-32 redundancy check code is used to ensure data integrity, further improving the reliability and response speed of the system.

[0045] In view of the problems of difficult thin plate welding deformation control and insufficient positioning accuracy in the prior art, the complete closed-loop control process is used in the embodiments of the present application, and high-precision clamping and deformation compensation are realized through multi-spectrum fusion positioning and self-adaptive pressing mechanism. In some embodiments of the present application, the central control unit 101 is further configured to, when the AGV intelligent transfer unit 102 transfers the workpiece to the assembly positioning station, project a precise positioning grid on the surface of the workpiece through a dual-wavelength laser, collect a grid image, extract feature corner points of the grid image, establish an initial deformation model of the cold van roof panel to be assembled by comparing the deviation of the theoretical coordinates and the actual coordinates of the feature corner points, and determine a displacement compensation amount according to real-time monitoring data of the distributed sensing network and the initial deformation model, and compensate the displacement of the workpiece through the displacement compensation amount. The real-time monitoring data includes material elastic modulus, strain gauge reading and environmental temperature change data.

[0046] The embodiments of the present application first perform initialization calibration, and the multi-spectrum laser-vision fusion positioning system enters a standby state. When the AGV intelligent transfer system transfers the workpiece (which can be a cold van roof panel (typical size: 3000mm*2000mm*1.5mm) to be assembled) to the assembly positioning station, the positioning system is immediately activated. The dual-wavelength laser (650nm and 850nm) projects an 8*8 precise positioning grid on the surface of the plate, and the industrial camera synchronously collects the grid image at a sampling rate of 120fps. The image processing module extracts 256 feature corner points from the collected image by using the SURF (SpeededUp Robust Features) algorithm, and establishes an initial deformation model of the plate by comparing the deviation of the theoretical coordinates and the actual coordinates, thereby providing a data basis for subsequent compensation.

[0047] Then in the deformation compensation calculation stage, the central control unit receives real-time data from the distributed sensor network, including the material elastic modulus (automatically retrieved according to the steel grade), the strain gauge reading (reflecting the internal stress state of the plate), and the environmental temperature change data collected by the temperature sensor. The system calculates the displacement compensation amount through the established deformation compensation model, which takes into account the interaction of material properties, thermal expansion effect and clamping force. The core calculation formula of the displacement compensation amount is: Δd = E × ε × Δt, where Δd represents the displacement compensation amount (unit: mm), E is the material elastic coefficient (determined by the steel grade, unit: GPa), ε is the strain gauge reading (dimensionless), and Δt represents the temperature change time (the time spent from the reference room temperature to the current temperature rise, unit: s). This formula quantifies the relationship between thermal stress and material elastic deformation, providing accurate compensation basis for adaptive compression.

[0048] The embodiment of the present application achieves sub-millimeter level active deformation control accuracy: by using a double closed-loop system of "multi-spectral laser-vision fusion positioning" and "adaptive compression mechanism", the clamping elastic deformation and welding thermal deformation of the plate are sensed and actively compensated in real time, the positioning reference plane flatness error is controlled within 0.05mm / m², and the contour accuracy and weld centering of the compartment body are fundamentally guaranteed.

[0049] In some embodiments of the present application, the central control unit 101 is further configured to, when the assembly positioning station cannot perform real-time compensation, divide the workpiece surface into a fine grid, obtain the curvature radius data of the center point of each grid through a three-dimensional scanner, calculate the curvature radius data according to a preset compression force optimization model, obtain an optimal compression force value, adjust the compression force of the plurality of compression heads through the optimal compression force value, ensure that each region of the workpiece is uniformly stressed and consistent with the theoretical curved surface, predict the rebound trend of the workpiece after the compression force is released based on a cold work hardening compensation algorithm, material mechanical property data and historical deformation records, obtain a reverse compensation force, and apply the reverse compensation force to the high-strength steel plate and then perform multiple iterative optimizations through an algorithm to obtain a compensated workpiece.

[0050] When the assembling and positioning station cannot perform real-time compensation, the embodiment of the present application performs adaptive compression, in the adaptive compression execution link, the system divides the surface of the workpiece (which can be a steel plate) into a fine grid of 50x50, and a three-dimensional scanner acquires the curvature radius data of the center point of each grid with a resolution of 0.1mm. The compression force optimization model calculates the optimal pressure value in real time according to the curvature distribution of the curved surface, and the calculation formula is: P_optimal=P_base+k×(1 / R)×f(T), wherein P_optimal represents the optimal pressure (unit: N), P_base is the basic pressure (to ensure that the plate is basically fixed), k is the material coefficient (dynamically adjusted according to the thickness of the steel plate), R is the radius of curvature (unit: m), and f(T) is the temperature influence index (reflecting the thermal expansion effect). The 80 independently controlled compression heads complete pressure adjustment within 0.5s, and the pressure adjustment range is 50N to 1500N, ensuring that the stress of each area of the plate is uniform and consistent with the theoretical curved surface. In order to eliminate the rebound effect of high-strength steel plates, the cold hardening compensation algorithm integrated by the system starts to work. Based on the material mechanics performance data and historical deformation records, the algorithm predicts the rebound trend of the plate after the compression force is released, and actively suppresses it by applying a reverse compensation force in advance. Through multiple iterations and optimization, the algorithm continuously corrects the compensation parameters, and finally stabilizes the flatness error of the positioning reference surface to within 0.05mm / m². During the entire positioning process, the system monitors the plate state in real time, and through feedback adjustment, ensures that the positioning accuracy always remains within the design requirement of ±0.1mm. In addition, the system realizes microsecond-level data synchronization through the time-sensitive network protocol, ensuring that there is no delay between the positioning instructions and the execution mechanism, thereby effectively solving the problem that traditional mechanical clamps cannot respond to material deformation in real time.

[0051] Further, the central control unit 101 is also used for controlling the low-temperature environment heating sub-unit 1004 to activate the ceramic heating film when in a low-temperature environment, and suppressing the interference of the welding electric arc in the gigahertz frequency band through the electromagnetic compatibility protection layer, and converting mechanical vibration into electrical energy and storing it in the buffer battery pack.

[0052] The embodiment of the present application provides a low-temperature environment heating sub-unit, which is used for activating the ceramic heating film when in a low-temperature environment (for example, -40 DEG C.), so as to keep the system temperature greater than 0 DEG C.; the electromagnetic compatibility protection layer suppresses the interference of the welding electric arc in the gigahertz frequency band on the control system; and the vibration energy recovery device converts mechanical vibration into electrical energy and stores it in the buffer battery pack.

[0053] Further, the distributed emergency stop module realizes 8ms-level response emergency braking through an optical fiber network; the dynamic electronic fence triggers the sound wave expeller when invading the warning area based on the millimeter wave radar to construct a personnel activity heat map; and the multi-physical field monitoring module analyzes temperature, vibration and electromagnetic radiation data in real time, and starts the workpiece pneumatic locking and system hibernation program when abnormal.

[0054] Enhanced system adaptability and safety under extreme conditions: integrated low-temperature environment protection, electromagnetic compatibility protection, vibration energy recovery and other extreme condition coping modules to ensure system stable operation in-40℃ high cold, strong electromagnetic interference and other harsh environments. At the same time, the multi-level safety protection system (distributed emergency stop, dynamic electronic fence, multi-physical field monitoring) provides all-round safety protection for personnel and equipment.

[0055] In order to better implement the tailor-welded workbench multi-station collaborative work system in the embodiment of the application, on the basis of the tailor-welded workbench multi-station collaborative work system, correspondingly, the embodiment of the application also provides a tailor-welded workbench multi-station collaborative work method, as shown in Figure 4 The tailor-welded workbench multi-station collaborative work method comprises: S401, when the tailor-welded workbench receives a task instruction, a task sequence distribution scheme is obtained based on a multi-station dynamic scheduling algorithm and the task instruction, and the corresponding task instruction is distributed to the multi-station according to the task sequence distribution scheme and a time-sensitive network protocol; S402, determining an optimal path according to the task instruction, and transferring workpieces between stations according to the optimal path.

[0056] The tailor-welded workbench multi-station collaborative work method provided in the above embodiment can realize the technical solutions described in the tailor-welded workbench multi-station collaborative work system embodiment. The principles of the implementation of the above modules or units can be referred to the corresponding content in the tailor-welded workbench multi-station collaborative work system embodiment, which will not be repeated here.

[0057] The tailor-welded workbench multi-station collaborative work system and method provided by the application are described in detail above, and the principles and implementation modes of the application are described in this paper. The above example is only used to help understand the method and its core idea; at the same time, for those skilled in the art, according to the idea of the application, the specific implementation mode and application range will be changed; in view of the above, the content of the specification should not be understood as a limitation of the application.

Claims

1. A tailor-welded-blank multi-station collaborative working system, characterized in that, The central control unit and the AGV intelligent transfer unit are included. The central control unit is configured to obtain a task sequence allocation scheme based on a multi-station dynamic scheduling algorithm and the task instruction when the tailor-welding workbench receives a task instruction, and distribute corresponding task instructions to the multi-station according to the task sequence allocation scheme and a time-sensitive network protocol. The AGV intelligent transfer unit is configured to determine an optimal path according to the task instruction, and transfer workpieces between stations according to the optimal path.

2. The tailor-welded-blank multi-station collaborative work system of claim 1, wherein, The multi-station includes interconnected automatic feeding stations, assembly positioning stations, multi-robot welding stations, and online detection stations.

3. The tailor-welded-blank multi-station collaborative work system of claim 1, wherein, The central control unit is further configured to perform multi-dimensional similarity calculation on a new vehicle model in the task instruction and vehicle models in a process knowledge base to obtain a best reference vehicle model with the maximum similarity, perform transfer learning on the best reference vehicle model to obtain initial process parameters, perform digital twinning on the initial process parameters to obtain a vehicle three-dimensional model, perform virtual debugging on the vehicle three-dimensional model to output an optimal fixture layout scheme and a robot motion trajectory, and perform dynamic scheduling on the optimal fixture layout scheme and the robot motion trajectory according to the multi-station dynamic scheduling algorithm to obtain a task sequence allocation scheme.

4. The multi-station collaborative operation system for welding workbench according to claim 1, characterized in that, The central control unit further includes determining corresponding fixture parameters and positioning instructions after each station receives the corresponding task instruction, changing the type of the fixture corresponding to the fixture parameters through a pneumatic-electromagnetic composite locking mechanism, and accurately positioning the fixture through the positioning instructions; the changing includes old fixture disassembly, new fixture installation, pose fine adjustment, and system self-checking.

5. The multi-station collaborative operation system for welding workbench according to claim 2, characterized in that, The central control unit includes a molten pool monitoring subunit; the central control unit is further configured to perform collaborative path planning according to the three-dimensional point cloud data of the weld of the workpiece and the dynamically reconfigurable topology layout formed by the multiple collaborative welding robots when the workpiece is welded at the multi-robot welding station, to obtain a welding area and an interference-free motion trajectory, and control the multiple collaborative welding robots to weld the workpiece according to the welding area and the interference-free motion trajectory, detect the molten pool oscillation frequency in real time through the molten pool monitoring subunit during the welding process, and perform defect early warning through the molten pool oscillation frequency, and optimize the welding parameters through a current-voltage-wire feeding speed mapping model.

6. The multi-station collaborative operation system for welding workbench according to claim 2, characterized in that, The ring guide rail of the assembly positioning station is provided with a hydraulic push rod of an elastic groove; the central control unit is further configured to start an elastic buffer adjustment mechanism when the assembly positioning station is expected to be delayed due to fixture replacement; the elastic buffer adjustment mechanism calculates buffer demand according to delay data, controls the hydraulic push rods of adjacent two elastic grooves on the ring guide rail to be started synchronously, and extends the groove length according to the buffer demand to form a temporary storage area, and buffers through the temporary storage.

7. The multi-station collaborative operation system for welding workbench according to claim 2, characterized in that, The central control unit is further configured to freeze the task queue of the abnormal station when the AGV intelligent transfer unit detects that the station abnormally stops, and reassign the uncompleted production tasks through the multi-station dynamic scheduling algorithm, and when the abnormal station is the multi-robot welding station, adjust the welding area division of each robot by dynamically reconstructing the robot topology layout, so that the AGV intelligent transfer unit re-plans the optimal path according to the reassigned tasks or the re-adjusted welding area.

8. The multi-station collaborative operation system for welding workbench according to claim 2, characterized in that, The central control unit is further configured to project a precision positioning grid on the surface of the workpiece by a dual-wavelength laser when the AGV intelligent transfer unit transports the workpiece to the assembly positioning station, and collect a grid image, extract feature corner points from the grid image, establish an initial deformation model of the workpiece by comparing the deviation of the theoretical coordinates and the actual coordinates of the feature corner points, and determine a displacement compensation amount according to the real-time monitoring data of the distributed sensing network and the initial deformation model, and compensate the displacement of the workpiece by the displacement compensation amount; the real-time monitoring data includes material elastic modulus, strain gauge reading and environmental temperature change data.

9. The multi-station collaborative operation system for welding workbench according to claim 8, characterized in that, The central control unit is further configured to divide the surface of the workpiece into a fine grid when the assembly positioning station cannot perform real-time compensation, obtain the curvature radius data of the center point of each grid by a three-dimensional scanner, calculate the curvature radius data according to a preset compression force optimization model to obtain an optimal pressure value, and adjust the pressure of multiple compression heads by the optimal pressure value to ensure that each region of the workpiece is uniformly stressed and consistent with the theoretical curved surface, and predict the springback trend of the workpiece after releasing the compression force based on a cold work hardening compensation algorithm, material mechanical property data and historical deformation records to obtain a reverse compensation force, and apply the reverse compensation force to the high-strength steel plate and then perform multiple iterative optimizations by an algorithm to obtain the compensated workpiece.

10. A method for multi-station collaborative work of a tailor-welded blank station, characterized in that, Comprise: When the tailor-welding workbench receives a task instruction, a task sequence allocation scheme is obtained based on a multi-station dynamic scheduling algorithm and the task instruction, and the corresponding task instruction is distributed to multiple stations according to the task sequence allocation scheme and a time-sensitive network protocol; An optimal path is determined according to the task instruction, and the workpiece is transferred between stations according to the optimal path.

Images