Real-time control method for dual-robot laser-electric arc hybrid welding based on digital twinning

Abstract

This invention belongs to the field of intelligent manufacturing and laser-arc hybrid welding technology, and discloses a real-time control method for dual-robot laser-arc hybrid welding based on digital twins. This method uses two robots and constructs a digital twin model of the laser-arc hybrid heat source, integrating a Gaussian distributed laser heat source and a double-ellipsoidal arc heat source to predict the dynamic behavior of the molten pool (molten pool flow, keyhole stability) and the probability of defect generation (porosity, cracks, etc.). Visual sensors and acoustic emission sensors on both sides of the welding robot capture the molten pool oscillation frequency and internal defect signals. A reinforcement learning algorithm is used to achieve adaptive compensation for welding path offset and adjustment of the laser-arc energy ratio. Closed-loop real-time control is formed by controlling the welding torch position and process parameters of the welding robot. Based on the digital twin model, the welding torches of the two robots can weld simultaneously on both sides and can also be independently adjusted according to the characteristics of specific materials while ensuring coupling, achieving high-quality welding.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing and laser-arc hybrid welding technology, and in particular to a real-time control method for dual-robot laser-arc hybrid welding based on digital twins. Background Technology

[0002] Laser-arc hybrid welding technology combines the high energy density and precision of laser welding with the deep penetration and adaptability of arc welding. It boasts advantages such as high speed, high efficiency, low deformation, and single-sided welding with double-sided forming, solving problems like the large heat-affected zone and high residual stress associated with traditional arc welding. Compared to single welding methods, laser-arc hybrid welding offers unique advantages. Its core mechanism lies in the synergistic effect of the laser and the arc, achieving control of the welding process through thermal field coupling and dynamic behavior regulation of the molten pool. However, in traditional laser-arc hybrid welding processes, the energy ratio of the laser and the arc is fixed, lacking the integration of digital twins for dynamic coordination throughout the entire process. This prevents real-time adjustments based on the dynamic behavior of the molten pool, resulting in low laser energy utilization and a high defect rate.

[0003] Double-sided welding balances the welding heat input through the synchronous action of heat sources on both sides, reducing deformation and residual stress caused by heat accumulation on one side. For example, in T-joints, synchronous double-sided welding results in a more uniform distribution of lateral shrinkage deformation and longitudinal inherent deformation, significantly improving stress concentration in the weld area. However, the synchronicity and coupling of double-sided welding has always been a challenge.

[0004] Digital twins establish virtual mappings of physical systems, combining real-time data with simulation models to achieve functions such as process optimization and predictive maintenance. In the welding field, their applications focus on welding process simulation, defect prediction, and parameter optimization. Currently, the integration of existing digital twin technology with welding technology mainly manifests in welding quality inspection and facilitating workers to find appropriate process parameters for efficient welding. However, it still retains a degree of non-autonomy, unable to rapidly adjust the welding process in real time based on digital twin models and algorithms, and unable to achieve dynamic collaborative control throughout the entire process. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention aims to provide a real-time collaborative control method for laser-arc hybrid welding processes based on digital twins, belonging to the fields of intelligent manufacturing and laser-arc hybrid welding technology. This method addresses the problems of insufficient energy coordination, fragmented process parameters, and high defect rates in traditional laser-arc hybrid welding. By constructing a cross-scale digital twin model, it deeply integrates the combined heat source characteristics of laser and arc, achieving adaptive control and process parameter optimization of the welding process. The system includes a combined heat source digital twin modeling module, a multiphysics field coupling simulation module, and a real-time collaborative control module. Specifically, the combined heat source modeling uses a Gaussian distributed laser heat source and a double-ellipsoidal arc heat source, combined with a quantum annealing algorithm to dynamically optimize the energy ratio of laser and arc, matching the molten pool's thermal requirements in real time. The multiphysics field simulation establishes a thermo-mechanical-fluid coupling model based on MARC software, capturing the molten pool oscillation frequency and internal defect signals through acoustic emission sensors and a terahertz imager, analyzing molten pool flow, keyhole stability, and solidification behavior, and predicting the probability of porosity and crack formation. Based on the received signal feedback, the real-time control module dynamically adjusts the welding path offset and energy parameters using a reinforcement learning algorithm, forming a closed-loop real-time control. This significantly improves the process stability and quality control capabilities of laser-arc hybrid welding, making it suitable for complex applications such as thin-walled aluminum alloys for aerospace and thick high-strength steel plates for shipbuilding. To achieve the above objectives, this invention adopts the following technical solution:

[0006] A real-time control method for dual-robot laser-arc hybrid welding based on digital twins, characterized by the following steps:

[0007] Step 1: Construct a cross-scale digital twin model and predict the microstructure evolution and macroscopic deformation of the weld seam through molecular dynamics and thermo-mechanical coupled finite element analysis;

[0008] Step 2: During the welding process, the visual sensors and acoustic emission sensors of the welding robot are used to acquire the defect signals inside the molten pool and the phase transformation characteristics of the material in real time, and the data is fed back to the computer in real time.

[0009] Step 3: Based on the federated learning framework and multi-site database, the computer forms a digital twin simulation comparison and matching relationship for the defect signals and digital twin prediction models, and optimizes the combination of process parameters through quantum annealing algorithm;

[0010] Step 4: The computer control system dynamically allocates welding tasks through the intelligent control cabinet, controls the welding laser, welding motor and welding robot to execute coordinated control instructions, and adopts the best welding process parameters in real time to achieve high-quality welding.

[0011] Furthermore, the digital twin model is established using SolidWorks professional 3D design software to construct a lightweight 3D model of the welded workpiece. The thermo-mechanical coupled finite element analysis technology employs a multiphysics simulation module, integrating the thermo-mechanical coupled model with a material library to calculate the dynamic behavior of the molten pool and welding deformation.

[0012] Furthermore, the multiphysics simulation module establishes a laser-arc composite heat source model. The laser heat source model uses a Gaussian distribution heat source, and the arc heat source model uses a double-ellipsoidal heat source model. In addition, by setting boundary conditions, the actual welding process can be simulated as comprehensively as possible, including various boundary conditions such as heat dissipation and displacement constraints. Welding process parameters include the laser focal diameter and the effective heating radius of the arc, and variables such as molten pool flow and residual stress distribution are calculated using a nonlinear solver and output to the digital twin database.

[0013] Furthermore, the establishment of the digital twin model includes the following steps:

[0014] SolidWorks was used to create a one-to-one model of the weldment and to mesh the model.

[0015] Add material properties. These properties include various characteristics such as mass density, Young's modulus, Poisson's ratio, specific heat capacity, and thermal conductivity. The more detailed the material properties are added, the higher the accuracy of the digital twin model.

[0016] Configure the welding path and weld seam settings, including the arc start point, arc end point, and welding direction.

[0017] Initial and boundary conditions need to be set. Initial conditions include ambient temperature, workpiece position, and initial stress. Boundary conditions need to be set according to the specific welding process. The heat source model for laser-arc hybrid welding typically uses a dual-heat source composite model with a Gaussian distributed heat source and a double ellipsoidal heat source. In addition to the heat source, boundary conditions that need to be added include heat dissipation coefficient, displacement constraint, and welding speed, which need to be flexibly set according to the actual situation.

[0018] The above conditions are checked, and if everything is correct, simulation is performed. The evolution of the weld microstructure and macroscopic deformation are predicted by molecular dynamics and thermo-mechanical coupled finite element analysis.

[0019] Furthermore, the visual sensor can detect the temperature distribution and thermal changes in the welding area in real time, and feed back the dynamic changes such as the depth, shape, and flow of the molten pool to the computer in real time; the acoustic emission sensor can analyze the acoustic emission signals during the welding process, and can detect and provide feedback on potential defects such as porosity, slag inclusions, and cracks.

[0020] Furthermore, after receiving the signal from the sensor, the computer can match the feedback defect signal with the digital twin model, and based on the prediction of the digital twin model, combined with the database and reinforcement learning algorithm, adjust the process parameters and robot motion trajectory.

[0021] Furthermore, the reinforcement learning algorithm adjusts process parameters such as laser power, arc current, and welding speed, and uses a search algorithm and welding database to select the optimal parameter combination to match the heat input requirements of the molten pool in real time. For T-joints, the molten pools formed on both sides during welding by the dual-robot process are not completely identical. Therefore, the process parameters used by the robots differ for different molten pool morphologies. It is necessary to compare with a digital twin model and use an optimization algorithm to obtain the process parameters for each robot, and adjust the process in real time. This can effectively suppress the formation of weld defects and promote effective fusion of materials.

[0022] Furthermore, the establishment of the process parameters and the adjustment of the motion trajectory of the dual-sided welding robot are specifically achieved through the following methods.

[0023] 1) Dynamic feature extraction of the molten pool. Based on the infrared thermal image sequence acquired by the visual sensor, the aspect ratio λ=W / L and temperature gradient ∇T of the molten pool are extracted; the energy entropy E is extracted from the acoustic emission signal as a defect probability index.

[0024] 2) Detection of the matching degree between the digital twin model and the actual process.

[0025] Define the difference function: D = α|λ1-λ0| + β|∇T1-∇T0| + γE

[0026] Where λ1 is the aspect ratio of the molten pool in the theoretical model, λ0 is the aspect ratio of the molten pool in the actual welding process, ∇T1 is the theoretical temperature gradient, and ∇T0 is the actual temperature gradient. E is used as a defect probability index; for example, when E = 1, defects will inevitably exist in the welding process. α, β, and γ are weighting coefficients, α + β + γ = 1. The proportion of each weight needs to be determined through material properties, principal component analysis, and actual welding process requirements. Generally, 0.04 ≤ α ≤ 0.05, β ≤ 0.01, and γ ≥ 0.95. Define D. normal When the difference function D ≤ 0.4, it can be determined that the actual welding process is basically consistent with the prediction of the digital twin model.

[0027] 3) Trigger parameter adjustment.

[0028] When the difference function D > 0.4, it is determined that there are significant deficiencies in the actual welding process, and the process parameters need to be adjusted promptly. At this point, a balance function is defined. R = k 1 D + k 2|Δ v|+ k 3 | Δ P 1∣+ k 4|Δ P 2 |

[0029] Where D is the difference value, Δ v Δ represents the difference in welding speed before and after adjustment. P 1 represents the difference in laser power before and after adjustment, Δ P 1 represents the difference between the arc power before and after adjustment. Arc power P=UI, which is the product of arc voltage and welding current. k 1, k 2, k 3 , k 4 represents the weighting coefficient. k 1+ k 2+ k 3+ k The weighting of each parameter (4=1) needs to be determined based on material properties, principal component analysis, and actual welding process requirements. Even for the same defect or to improve the weld pool, there are multiple adjustment methods, and various process parameters can change synergistically. Therefore, using a balance function can limit the adjustment range of each process parameter, thereby avoiding new welding problems caused by excessive parameter changes. During parameter adjustment, a reinforcement learning algorithm is first used to provide multiple improvement schemes based on the welding database. Then, the parameter fluctuation values ​​in each scheme are introduced into the balance function, resulting in multiple R values. When the R value of the balance function is minimized, the corresponding scheme can be considered the optimal solution.

[0030] 4) Robotic arm trajectory correction.

[0031] The formula for compensating the motion trajectory of a robotic arm is: Δ(x, y, z) = δ·cosθ(x, y, z) + K p ∫δdt

[0032] Where δ is the real-time offset between the center of the molten pool and the theoretical position of the weld (measured by a vision sensor), θ(x, y, z) is the angle between the welding direction and the X, Y, Z axes of the robotic arm coordinate system, and K p The integral gain coefficient determines the strength of the correction for historical errors. ∫δdt represents the cumulative effect of the offset over time and is used to quantify long-term systematic errors. δ⋅cosθ(x, y, z) can be used to calculate the lateral compensation (instantaneous deviation in fast response) based on the current molten pool offset δ and the welding direction angle θ.

[0033] Furthermore, the computer control system combines the digital twin simulation model and prediction results, detection feedback signals, and optimized process parameters and welding trajectory correction parameters to issue control commands to systems such as the laser generator, welding power supply, welding robot, and wire feeder. The systems cooperate and adjust in real time to ultimately achieve closed-loop control in the welding process and complete high-quality and efficient adaptive welding.

[0034] The advantages and benefits of this invention are as follows:

[0035] This invention provides a real-time control method for dual-robot laser-arc hybrid welding based on digital twins. By constructing a cross-scale digital twin model, it achieves multi-physics coupling simulation of microscopic molten pool flow and macroscopic weld formation. Combined with real-time feedback from visual and acoustic emission signals, it actively suppresses typical defects in hybrid welding such as porosity and cracks. Simultaneously, addressing the issue of mismatched process parameters on both sides during T-joint dual-robot welding, it utilizes digital twin technology to achieve interaction between welding process parameters and model simulation, enabling real-time control of process parameters such as adaptive compensation of the welding path, adjustment of the laser-arc energy ratio, and welding speed. This invention can achieve adaptive closed-loop control, providing a fully software-defined intelligent solution for precision welding in aerospace, new energy vehicles, and other fields, and has broad application prospects. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of a real-time control method for a dual-robot laser-arc hybrid welding process based on digital twins;

[0037] Figure 2 It is a digital twin platform for real-time control of a dual-robot laser-arc hybrid welding process based on digital twins;

[0038] Figure 3 This is a flowchart of real-time control of a dual-robot laser-arc hybrid welding process based on digital twins;

[0039] Figure 4 This is an example diagram of robot welding on both sides of an aluminum alloy T-joint. Detailed Implementation

[0040] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not the entire structure.

[0041] In the description of this invention, unless otherwise explicitly specified and limited, the terms "connected," "linked," and "fixed" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0042] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0043] In the description of this embodiment, the terms "upper," "lower," "right," etc., refer to the orientation or positional relationship shown in the accompanying drawings. They are used only for ease of description and simplification of operation, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the present invention. In addition, the terms "first" and "second" are used only for distinction in description and have no special meaning.

[0044] The present invention is further illustrated below by a specific embodiment. This embodiment describes a laser-arc hybrid welding process for a frame-truss rocket propellant tank wall panel structure. The rocket propellant tank wall panel is a T-shaped structure, and the material used is 2219 aluminum alloy.

[0045] A real-time control method for dual-robot laser-arc hybrid welding based on digital twins, such as... Figure 1 As shown, it mainly includes 1) a robotic welding system, 2) a digital twin simulation system, 3) a monitoring system (including visual sensors and acoustic emission sensors), and 4) a control system.

[0046] Preferably, such as Figure 2 As shown, a real-time control method for dual-robot laser-arc hybrid welding based on digital twins is implemented through the following steps:

[0047] Step 1: Construct a cross-scale digital twin model and predict the microstructure evolution and macroscopic deformation of the weld seam through molecular dynamics and thermo-mechanical coupled finite element analysis;

[0048] Step 2: During the welding process, the visual sensors and acoustic emission sensors of the welding robot are used to acquire the defect signals inside the molten pool and the phase transformation characteristics of the material in real time, and the data is fed back to the computer in real time.

[0049] Step 3: Based on the federated learning framework and multi-site database, the computer forms a digital twin simulation comparison and matching relationship for the defect signals and digital twin prediction models, and optimizes the combination of process parameters through quantum annealing algorithm;

[0050] Step 4: The computer control system dynamically allocates welding tasks through the intelligent control cabinet, controls the welding laser, welding motor and welding robot to execute coordinated control instructions, and adopts the best welding process parameters in real time to achieve high-quality welding.

[0051] Preferably, the digital twin model is established using SolidWorks professional 3D design software to construct a lightweight 3D model of the rocket propellant tank wall structure. The thermo-mechanical coupled finite element analysis technology employs a multiphysics simulation module, integrating the thermo-mechanical coupled model with a material library to calculate the dynamic behavior of the molten pool and welding deformation.

[0052] Preferably, the multiphysics simulation module establishes a laser-arc composite heat source model, wherein the laser heat source model adopts a Gaussian distributed heat source, and the arc heat source model adopts a double ellipsoidal heat source model. Furthermore, by setting boundary conditions, the actual welding process can be simulated as comprehensively as possible, including various boundary conditions such as heat dissipation and displacement constraints. Welding process parameters include laser focal diameter, effective arc heating radius, etc., and variables such as molten pool flow and residual stress distribution are calculated using a nonlinear solver and output to the digital twin database.

[0053] Preferably, the establishment of the digital twin model includes the following steps:

[0054] 1) Use SolidWorks to create a one-to-one model of the weldment and mesh the model.

[0055] 2) Add material properties. These properties include various characteristics such as mass density, Young's modulus, Poisson's ratio, specific heat capacity, and thermal conductivity. The more detailed the material properties are, the higher the accuracy of the digital twin model.

[0056] 3) Set the welding path and weld seam settings. This includes the arc start point, arc end point, and welding direction. For the double-sided welding of the T-shaped structure of the rocket propellant tank wall, two weld seams and two welding paths are required.

[0057] 4) Set initial and boundary conditions. Initial conditions include ambient temperature, workpiece position, and initial stress. Boundary conditions need to be set according to the specific welding process. The heat source model for laser-arc hybrid welding typically uses a dual-heat source composite model with a Gaussian distributed heat source and a double ellipsoidal heat source. In addition to the heat source, boundary conditions that need to be added include heat dissipation coefficient, displacement constraint, and welding speed, which need to be flexibly set according to the actual situation.

[0058] 5) Check the above conditions. If everything is correct, perform simulation and predict the microstructure evolution and macroscopic deformation of the weld through molecular dynamics and thermo-mechanical coupled finite element analysis.

[0059] Preferably, the visual sensor can detect the temperature distribution and thermal changes in the welding area in real time, and feed back the dynamic changes such as the depth, shape, and flow of the molten pool to the computer in real time; the acoustic emission sensor can analyze the acoustic emission signals during the welding process, and can detect and provide feedback on potential defects such as porosity, slag inclusions, and cracks.

[0060] Preferably, after receiving the signal from the sensor, the computer can match the feedback defect signal with the digital twin model, and based on the prediction of the digital twin model, combined with the database and reinforcement learning algorithm, adjust the process parameters and robot motion trajectory.

[0061] Preferably, the reinforcement learning algorithm adjusts process parameters such as laser power, arc current, and welding speed, and filters the optimal parameter combination through a search algorithm and welding database to match the heat input requirements of the molten pool in real time. For T-joints, the molten pools formed on both sides during welding by the dual-robot process are not completely identical. Therefore, the process parameters used by the robots are not the same for different molten pool morphologies. It is necessary to compare with the digital twin model and use optimization algorithms to obtain the process parameters of each robot on both sides, and adjust the process in real time. This can effectively suppress the formation of weld defects and promote effective fusion of materials.

[0062] Preferably, the establishment of the process parameters and the adjustment of the motion trajectory of the dual-sided welding robot are achieved in the following ways.

[0063] 1) Dynamic feature extraction of the molten pool. Based on the infrared thermal image sequence acquired by the visual sensor, the aspect ratio λ=W / L and temperature gradient ∇T of the molten pool are extracted; the energy entropy E is extracted from the acoustic emission signal as a defect probability index.

[0064] 2) Detection of the matching degree between the digital twin model and the actual process.

[0065] Define the difference function: D = α|λ1-λ0| + β|∇T1-∇T0| + γE

[0066] Where λ1 is the aspect ratio of the molten pool in the theoretical model, λ0 is the aspect ratio of the molten pool in the actual welding process, ∇T1 is the theoretical temperature gradient, and ∇T0 is the actual temperature gradient. E is used as a defect probability index; for example, when E = 1, defects will inevitably exist in the welding process. α, β, and γ are weighting coefficients, α + β + γ = 1. The proportion of each weight needs to be determined through material properties, principal component analysis, and actual welding process requirements. Generally, 0.04 ≤ α ≤ 0.05, β ≤ 0.01, and γ ≥ 0.95. Define D. normal When the difference function D ≤ 0.4, it can be determined that the actual welding process is basically consistent with the prediction of the digital twin model.

[0067] 3) Trigger parameter adjustment.

[0068] When the difference function D > 0.4, it is determined that there are significant deficiencies in the actual welding process, and the process parameters need to be adjusted promptly. At this point, a balance function is defined. R = k 1 D + k 2|Δ v |+ k 3 | Δ P 1∣+ k 4|Δ P 2 |

[0069] Where D is the difference value, Δ v Δ represents the difference in welding speed before and after adjustment. P 1 represents the difference in laser power before and after adjustment, Δ P 1 represents the difference between the arc power before and after adjustment. Arc power P=UI, which is the product of arc voltage and welding current. k 1, k 2, k 3 , k 4 represents the weighting coefficient. k 1+ k 2+ k 3+ k The weighting of each parameter (4=1) needs to be determined based on material properties, principal component analysis, and actual welding process requirements. Even for the same defect or to improve the weld pool, there are multiple adjustment methods, and various process parameters can change synergistically. Therefore, using a balance function can limit the adjustment range of each process parameter, thereby avoiding new welding problems caused by excessive parameter changes. During parameter adjustment, a reinforcement learning algorithm is first used to provide multiple improvement schemes based on the welding database. Then, the parameter fluctuation values ​​in each scheme are introduced into the balance function, resulting in multiple R values. When the R value of the balance function is minimized, the corresponding scheme can be considered the optimal solution.

[0070] 4) Robotic arm trajectory correction.

[0071] The formula for compensating the motion trajectory of a robotic arm is: Δ(x, y, z) = δ·cosθ(x, y, z) + K p ∫δdt

[0072] Where δ is the real-time offset between the center of the molten pool and the theoretical position of the weld (measured by a vision sensor), θ(x, y, z) is the angle between the welding direction and the X, Y, Z axes of the robotic arm coordinate system, and K p The integral gain coefficient determines the strength of the correction for historical errors. ∫δdt represents the cumulative effect of the offset over time and is used to quantify long-term systematic errors. δ⋅cosθ(x, y, z) can be used to calculate the lateral compensation (instantaneous deviation in fast response) based on the current molten pool offset δ and the welding direction angle θ.

[0073] Preferably, the computer control system combines the digital twin simulation model and prediction results, detection feedback signals, and optimized process parameters and welding trajectory correction parameters to issue control commands to systems such as the laser generator, welding power supply, welding robot, and wire feeder. The systems cooperate and adjust in real time to ultimately achieve closed-loop control in the welding process and complete high-quality and efficient adaptive welding.

[0074] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will be able to make various obvious changes, readjustments, and substitutions without departing from the scope of protection of the present invention. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A real-time control method for dual-robot laser-arc hybrid welding based on digital twins, characterized in that, Specifically, the steps include the following: Step 1: Construct a cross-scale digital twin model and predict the microstructure evolution and macroscopic deformation of the weld seam through molecular dynamics and thermo-mechanical coupled finite element analysis; Step 2: During the welding process, the visual sensors and acoustic emission sensors of the welding robot are used to acquire the defect signals inside the molten pool and the phase transformation characteristics of the material in real time, and the data is fed back to the computer in real time. Step 3: Based on the federated learning framework and multi-site database, a digital twin simulation comparison and matching relationship is formed for the defect signals and digital twin prediction models, and the combination of process parameters is optimized through quantum annealing algorithm; Step 4: The control system dynamically allocates welding tasks through the intelligent control cabinet, controls the welding laser, welding motor and welding robot to execute coordinated control instructions, and adopts the best welding process parameters in real time to achieve high-quality welding; The method further includes: After receiving the signal from the sensor, the computer matches the feedback defect signal with the digital twin model. Based on the prediction of the digital twin model, combined with the database and reinforcement learning algorithm, it adjusts the process parameters and the robot's motion trajectory. The reinforcement learning algorithm is specifically as follows: The system is designed to perform welding on T-joints. Two robots perform welding on both sides. The process parameters of laser power, arc current and welding speed are adjusted. The optimal parameter combination is selected by searching the algorithm and welding database to obtain the process parameters of each robot. The establishment of process parameters and the adjustment of the motion trajectory of the double-sided welding robot are achieved in the following ways: 1) Based on the infrared thermal image sequence acquired by the visual sensor, the aspect ratio λ=W / L of the molten pool and the temperature gradient ∇T are extracted; the energy entropy E of the acoustic emission signal is extracted as a defect probability index. 2) Define the difference function: D = α|λ1-λ0| + β|∇T1-∇T0| + γE; Where λ1 is the aspect ratio of the molten pool in the theoretical model, λ0 is the aspect ratio of the molten pool in the actual welding process, ∇T1 is the theoretical temperature gradient, and ∇T0 is the actual temperature gradient; E is the defect probability index, α, β, and γ are weighting coefficients, α+β+γ=1, and the proportion of each weight needs to be determined through material properties, principal component analysis, and actual welding process requirements. 3) When the difference function D > the set value, it is determined that there are obvious deficiencies in the actual welding process, and the process parameters need to be adjusted in time. In this case, a balance function is defined. R = k 1 D + k 2|Δ v |+ k 3 | Δ P 1∣+ k 4|Δ P 2 |; Where D is the difference value, Δ v Δ represents the difference in welding speed before and after adjustment. P 1 represents the difference in laser power before and after adjustment, Δ P 2 The difference between the arc power before and after adjustment is given by arc power P=UI, which is the product of arc voltage and welding current. k 1, k 2, k 3, k 4 represents the weighting coefficient. k 1+ k 2+ k 3+ k 4=1, and the weight ratio of each component needs to be determined based on material properties, principal component analysis, and actual welding process requirements; During the parameter adjustment process, firstly, multiple improvement schemes are provided based on the welding database using reinforcement learning algorithms. Then, the parameter fluctuation values ​​in each scheme are introduced into the balance function to obtain multiple balance function R values. When the balance function R value is the smallest, the corresponding scheme is taken as the optimal solution. 4) Set the formula for compensating the motion trajectory of the robotic arm as: Δ(x, y, z) = δ·cosθ(x, y, z) + K p ∫δdt; Where δ is the real-time offset between the center of the molten pool and the theoretical position of the weld, θ(x, y, z) is the angle between the welding direction and the X, Y, Z axes of the robotic arm coordinate system, and K p The integral gain coefficient determines the strength of the correction for historical errors. ∫δdt represents the cumulative effect of the offset over time and is used to quantify long-term systematic errors. δ⋅cosθ(x, y, z) can calculate the lateral compensation based on the current molten pool offset δ and the welding direction angle θ.

2. The real-time control method for dual-robot laser-arc hybrid welding based on digital twins according to claim 1, characterized in that, The digital twin model was established using SolidWorks professional 3D design software to construct a lightweight 3D model of the welded workpiece; the thermo-mechanical coupled finite element analysis technology adopted a multiphysics simulation module, integrating the thermo-mechanical coupled model and material library, to calculate the dynamic behavior of the molten pool and welding deformation.

3. The real-time control method for dual-robot laser-arc hybrid welding based on digital twins according to claim 2, characterized in that, The multiphysics simulation module establishes a laser-arc composite heat source model, in which the laser heat source model adopts a Gaussian distributed heat source, and the arc heat source model adopts a double ellipsoidal heat source model. In addition, by setting boundary conditions, the actual welding process can be simulated as comprehensively as possible, including various boundary conditions such as heat dissipation and displacement constraints. The welding process parameters include the laser focus diameter and the effective heating radius of the arc, and the parameters of molten pool flow and residual stress distribution variables are calculated by a nonlinear solver and output to the digital twin database.

4. The real-time control method for dual-robot laser-arc hybrid welding based on digital twins according to claim 1, characterized in that, The visual sensor can detect the temperature distribution and thermal changes in the welding area in real time, and dynamically feed back the changes in the depth, shape and flow of the molten pool to the computer in real time; the acoustic emission sensor can analyze the acoustic emission signals during the welding process, and can detect and provide feedback on potential defects such as porosity, slag inclusions and cracks.

5. The real-time control method for dual-robot laser-arc hybrid welding based on digital twins according to claim 1, characterized in that, The control system combines the digital twin simulation model and prediction results, detection feedback signals, and optimized process parameters and welding trajectory correction parameters to issue control commands to the laser generator, welding power source, welding robot, and wire feeder system.

Images