Gate welding intelligent obstacle avoidance method and system based on digital twinning technology

Abstract

The application is suitable for the technical field of gate welding, and provides a gate welding intelligent obstacle avoidance method and system based on digital twinning technology, which comprises the following steps: constructing a full-element digital twinning model containing geometric, material, assembly and process information; performing collision avoidance pre-simulation based on the model, and generating a collision-free welding path through iterative optimization; perceiving static and dynamic obstacles in real time through visual, distance and force sensing multi-source sensors during the welding process; mapping the perceived data to the digital twinning model to realize virtual-real synchronization, generating an obstacle avoidance adjustment strategy when a collision risk is detected or predicted, dynamically optimizing the welding torch trajectory and welding parameters, and smoothly returning to the original path after obstacle avoidance; feeding the obstacle avoidance, path adjustment and quality detection data to the model to update the obstacle avoidance strategy library and form a closed-loop optimization system. The application realizes full-process digital closed-loop control and dynamic obstacle avoidance of the gate welding process, and effectively improves the welding safety, quality stability and production efficiency.

Description

Technical Field

[0001] This invention belongs to the field of gate welding technology, and in particular relates to an intelligent obstacle avoidance method and system for gate welding based on digital twin technology, which is especially suitable for path planning, real-time obstacle avoidance and closed-loop optimization control in complex welding operation scenarios of large water conservancy gates. Background Technology

[0002] Hydraulic gates are core load-bearing components of water conservancy projects. The gate bodies are mostly large, thick-plate welded steel structures, characterized by large dimensions, numerous welds, and complex weld spatial arrangements. After on-site assembly, tooling fixtures, supporting components, and surrounding temporary facilities are interspersed, significantly limiting the welding work space. With the development of intelligent manufacturing technology, nine-axis linkage welding robots are gradually being applied to the automated welding operations of large gates. These robots, with their multi-axis motion capabilities, can achieve accessible welding of welds in complex spaces. However, in actual welding production, they still face severe collision and interference challenges.

[0003] Obstacles in large gate welding scenarios are categorized into static and dynamic obstacles. Static obstacles include gate support fixtures, clamping jigs, welding bases, and fixed workshop equipment. Dynamic obstacles include the thermal deformation profile of components after welding, mobile workers on-site, and mobile auxiliary fixtures. Traditional robotic welding paths mostly rely on offline programming for path planning. Offline programming can only perform path deduction based on the design model, considering only the spatial position under the theoretical model. It cannot take into account on-site assembly deviations and contour changes caused by welding thermal deformation. Furthermore, it lacks real-time perception and prediction capabilities for dynamic obstacles. During actual gate welding operations, the welding torch, robot body, and auxiliary axes are highly susceptible to collisions and interference with gate components and fixtures. Collisions can cause minor damage to the welding torch and weld trajectory deviation, leading to welding defects such as incomplete penetration and undercut, reducing gate welding quality. More serious collisions can damage the expensive nine-axis linkage welding robot body, causing component impact damage, resulting in significant economic losses and severely impacting gate production progress.

[0004] Existing obstacle avoidance solutions for robotic welding mainly fall into two categories. One category relies solely on visual and force sensors for passive obstacle avoidance after contact. This type of solution is a post-collision response, only capable of stopping and retracting after a collision, lacking a pre-simulation prediction mechanism. This leaves the system with little time to react, making it highly susceptible to hardware contact damage. The other category is a pre-obstacle avoidance solution based on offline simulation. This solution only performs static path verification in offline software, lacking real-time bidirectional data mapping between the physical entity and the virtual model. It cannot perceive dynamic obstacles during the welding process and cannot dynamically replan the welding path based on real-time on-site conditions.

[0005] Further searching existing patent literature revealed that CN120421824A discloses an intelligent welding control method for gates. This solution focuses on controlling the reverse disturbance of abnormal welding stress but does not address the collision avoidance problem between moving parts of the welding equipment and surrounding components. CN120696655A emphasizes the improvement of the mechanical structure of the welding device but does not combine digital twins to achieve a complete technical system of pre-simulation—real-time perception—dynamic obstacle avoidance—closed-loop iteration. Most of the existing publicly disclosed digital twin welding-related technologies focus on the simulation of welding temperature field and stress field, and the optimization of welding process parameters. There are few disclosures of integrated obstacle avoidance technology solutions for complex gate operating conditions, which combine "offline pre-simulation to avoid static interference + online multi-source sensing to perceive dynamic obstacles + digital twin virtual-real synchronous dynamic replanning".

[0006] In summary, the existing technology has the following drawbacks: First, offline path planning is disconnected from actual on-site working conditions. Collision detection is based solely on an ideal 3D model and cannot adapt to the contour changes caused by gate assembly gap deviations and welding thermal deformation. Second, passive sensor obstacle avoidance is an emergency response after a collision occurs, lacking the ability to predict the movement trend of obstacles, making it difficult to provide early warning of collision risks. Third, the virtual simulation model and the physical welding entity lack a high-frequency virtual-real synchronization mechanism, and the pre-simulation results cannot be dynamically updated based on the on-site perception data. Fourth, the lack of a closed-loop iterative optimization mechanism means that historical obstacle avoidance cases cannot be accumulated and reused, and the path needs to be re-adjusted for each welding task, resulting in low production debugging efficiency. Fifth, the kinematic constraints of the multiple additional axes of the nine-axis linkage welding robot were not jointly verified, which may easily lead to problems such as the joints and additional axes exceeding the motion limits and thus failing to execute, even though there is no interference in the path space.

[0007] Therefore, there is an urgent need to develop an intelligent obstacle avoidance method for gate welding based on digital twin technology, which integrates offline pre-simulation collision detection, real-time perception of multi-source sensors, virtual-real synchronous mapping, collision trend prediction, and closed-loop strategy iterative update to solve the problem of static and dynamic obstacle collision interference in the welding process of large gate robots, ensure the safety of welding equipment, and at the same time guarantee the quality of weld seams. Summary of the Invention

[0008] The purpose of this invention is to provide an intelligent obstacle avoidance method and system for gate welding based on digital twin technology, which aims to solve the problems of lack of real-time response in path planning, single obstacle avoidance strategy, and disconnect between design and execution in traditional welding.

[0009] This invention is implemented as follows: a gate welding intelligent obstacle avoidance method based on digital twin technology, relying on 3D modeling software, a finite element simulation platform, a nine-axis linkage welding robot, and a multi-source sensor acquisition system to realize gate welding operation path planning and obstacle avoidance; the gate welding intelligent obstacle avoidance method includes the following steps: S1. Collect the three-dimensional design data of the gate components and establish a full-element digital twin model containing geometric structure information, material property information, assembly relationship information and welding process parameter information; based on the kinematic parameters of the welding equipment, construct a virtual welding equipment model and a virtual working environment model that correspond one-to-one with the physical welding equipment in the simulation platform. The virtual welding equipment model includes at least a welding robot body model, an additional axis model and a welding torch model. S2. Based on the full-element digital twin model, automatically extract weld feature information to generate the initial welding path and initial welding torch posture; drive the virtual welding equipment model to move along the initial welding path in the simulation platform, perform collision avoidance simulation analysis, calculate the spatial distance between each moving part of the welding equipment and the gate component and surrounding obstacles in real time, and determine that there is a collision risk when any spatial distance is less than the preset safety threshold, triggering an automatic adjustment strategy. The automatic adjustment strategy includes at least one of path offset adjustment, posture rotation adjustment and welding sequence adjustment, and generates a collision-free welding path and corresponding obstacle avoidance process scheme through iterative optimization. S3. During the welding process, visual sensors, distance sensors, and force sensors installed on the welding equipment collect real-time perception data of the welding operation area. Based on the perception data, static obstacle information and dynamic obstacle information are identified, and the relative pose relationship between the welding torch and each obstacle is calculated in real time. S4. The real-time sensing data is mapped to a full-element digital twin model after coordinate transformation, so that the pose of the virtual welding equipment is synchronized with the actual pose of the physical welding equipment. Based on the relative pose relationship, when an obstacle is detected to enter the preset safe distance range or a collision is predicted, the digital twin model generates an obstacle avoidance adjustment strategy, dynamically adjusts the welding torch trajectory and / or dynamically adjusts the welding parameters, controls the welding equipment to perform obstacle avoidance actions, and automatically returns to the planned position of the preset welding path after obstacle avoidance is completed. S5. Feed back obstacle avoidance data, path adjustment data and welding quality inspection data generated during the actual welding process to the full-element digital twin model, update the model parameters and obstacle avoidance strategy library, and form a closed-loop optimization system of "modeling-planning-perception-obstacle avoidance-feedback".

[0010] A further technical solution of the present invention is as follows: In step S1, the full-element digital twin model establishes a modular and parametric three-dimensional geometric model of the gate component based on three-dimensional modeling software. The geometric structure information includes the weld groove size, plate thickness, spatial geometric contour, and assembly gap between components of the gate component; the material property information includes the thermal conductivity, specific heat capacity, yield strength, and weldability parameters of commonly used steel for gates; the assembly relationship information includes the connection method, fit tolerance, and assembly sequence between each component; and the welding process parameter information includes the welding current range, welding voltage range, welding speed range, and welding torch angle range corresponding to different weld types. The virtual working environment model includes three-dimensional spatial occupancy information of fixed obstacles around the gate component, spatial occupancy information of temporary facilities, and workspace boundary information of the welding equipment; the kinematic model of the welding robot body in the virtual welding equipment model is established using the Denavit-Hartenberg parametric method, the additional axis model includes kinematic parameters of seven to nine additional axes, and the mapping accuracy between the virtual welding equipment model and the physical welding equipment is not less than ±0.5mm.

[0011] A further technical solution of the present invention is: the collision avoidance simulation analysis in step S2 adopts a hierarchical bounding box collision detection algorithm, which performs hierarchical intersection tests on the bounding box hierarchical structure of each moving part of the welding equipment and obstacles, and calculates the minimum distance between each part; when the minimum distance is less than a preset safety threshold, the collision area is marked and the welding equipment pose parameters at the moment of collision are recorded; the preset safety threshold is 5mm to 20mm, and different safety thresholds are set for different types of obstacles; The path offset adjustment includes: keeping the welding torch posture unchanged, and offsetting and correcting the path coordinates in planes perpendicular to and parallel to the welding forward direction, with the deviation after path correction controlled within ±0.5mm; the posture rotation adjustment includes: keeping the spatial coordinates of the welding path unchanged, adjusting the pitch angle, yaw angle, and roll angle of the welding torch, and optimizing the welding torch posture through additional axis linkage to avoid obstacles, with the angle deviation between the welding torch axis and the weld centerline ≤±1° after adjustment; the welding sequence adjustment includes: adjusting the welding sequence of weld segments with collision risks, and adjusting high-risk areas to a state of no obstacles for welding.

[0012] A further technical solution of the present invention is as follows: the vision sensor in step S3 includes a 3D structured light vision sensor and a laser contour scanning sensor. The 3D structured light vision sensor is used to collect three-dimensional point cloud data of the welding operation area, and the laser contour scanning sensor is used to collect two-dimensional cross-sectional data of the weld contour. The distance sensor includes an ultrasonic sensor and a laser ranging sensor, which are used for mid-to-long-range obstacle detection and close-range precise ranging, respectively, with a measurement accuracy of not less than ±0.5mm. The force sensor is a six-dimensional force / torque sensor, which is set at the connection flange between the welding torch and the welding robot body, and is used to collect three-dimensional force and three-dimensional torque data when the welding torch contacts the obstacle in real time, with a sampling frequency of not less than 1000Hz. The static obstacle information includes the spatial position and geometric dimensions of the welding equipment mounting base, gate support fixture, clamps, and surrounding fixed equipment; the dynamic obstacle information includes the dynamic contour changes caused by thermal deformation of the gate components during welding, the real-time position information of personnel moving around the welding equipment, and the real-time position information of mobile devices; the method for identifying the dynamic obstacle information is as follows: a real-time three-dimensional point cloud model of the welding operation area is constructed based on the real-time three-dimensional point cloud data collected by the visual sensor; the real-time three-dimensional point cloud model is registered and differentially compared with the full-element digital twin model; the difference regions are extracted and clustered to segment the difference regions, thereby identifying the spatial position and contour information of each dynamic obstacle.

[0013] A further technical solution of the present invention is as follows: the step S4 of mapping the real-time sensing data to the full-element digital twin model after coordinate transformation includes: establishing a coordinate transformation matrix between the sensor coordinate system and the welding robot base coordinate system, transforming the sensing data collected by the vision sensor, distance sensor and force sensor to a unified global coordinate system; and mapping the data to the corresponding virtual device and virtual environment model in the simulation platform after demodulation and filtering, so that the pose of the virtual welding device is synchronized with the actual pose of the physical welding device, the virtual and actual pose synchronization frequency is not less than 10Hz, and the pose synchronization delay is not more than 50ms. The method for predicting an impending collision is as follows: based on the real-time relative pose between the welding torch and the obstacle, and the current speed and acceleration of the welding equipment, a Kalman filter algorithm is used to predict the trajectory of the welding torch within a preset time window. The predicted minimum distance between the predicted trajectory and the space occupied by the obstacle is calculated. When the predicted minimum distance is less than a preset safety threshold, it is determined that a collision is imminent. The preset time window is 500ms to 2000ms, and the preset safety threshold is 10mm to 30mm. The safety threshold is dynamically adjusted according to the speed of the welding equipment; the higher the speed, the larger the safety threshold. In step S4, the three levels of interferometry early warning are determined based on the predicted minimum distance: Level 1 slight deviation, meaning the predicted minimum distance is greater than the first preset safety threshold, only the deviation data is recorded; Level 2 moderate interference risk, i.e. the predicted minimum distance is greater than 0 and less than or equal to the first preset safety threshold, output trajectory slow fine-tuning instruction; Level 3 severe interference risk, i.e. the predicted minimum distance is less than or equal to 0, output emergency stop and welding torch retraction command; The obstacle avoidance adjustment strategy is a combination of one or more of the following strategies: a) Deceleration strategy: Reduce the speed of the welding equipment to below the preset safe speed, and the deceleration time shall not exceed 100ms; b) Stopping strategy: Control the welding equipment to stop its movement in an emergency, with a braking distance not exceeding 5mm; c) Offset strategy: Control the welding torch to offset along the normal direction of the obstacle surface by a preset offset amount to avoid the obstacle. The offset amount is dynamically calculated based on the geometric dimensions of the obstacle. d) Attitude adjustment strategy: By adjusting the pitch and yaw angles of the welding torch, the spatial orientation of the welding torch can be changed, so that the welding torch can avoid obstacles while maintaining welding operation capability; The priority order of the obstacle avoidance adjustment strategies is as follows: a combination of deceleration strategy and attitude adjustment strategy is preferred; when attitude adjustment cannot meet the obstacle avoidance requirements, an offset strategy is activated; and when the offset strategy still cannot avoid the obstacle, a stopping strategy is activated. The dynamic adjustment of welding parameters in step S4 includes: during obstacle avoidance, dynamically adjusting the welding current and welding voltage based on the relative posture deviation between the welding torch and the weld to maintain a stable heat input and ensure the consistency of weld quality between the obstacle avoidance section and the non-obstacle avoidance section; after obstacle avoidance is completed, the digital twin model generates a transition path that smoothly returns from the current position to the preset welding path through a path interpolation algorithm, guiding the welding equipment to return to the planned position of the preset welding path along the transition path. The curvature of the transition path is continuous, and the rate of change of acceleration does not exceed a preset comfort threshold.

[0014] A further technical solution of the present invention is as follows: the obstacle avoidance data in step S5 includes the timestamp of the obstacle avoidance trigger time, the three-dimensional coordinates of the obstacle avoidance trigger position, the path trajectory point sequence during the obstacle avoidance execution process, the welding parameter adjustment record during the obstacle avoidance process, and the timestamp of the obstacle avoidance completion time; the path adjustment data includes the deviation data and deviation correction amount between the initial welding path and the actual execution path after obstacle avoidance; the welding quality inspection data includes weld forming size inspection data and weld internal defect inspection data; the updating of model parameters and obstacle avoidance strategy library specifically includes: performing cluster analysis on multiple obstacle avoidance data through machine learning algorithms, identifying high-frequency obstacle avoidance areas and high-frequency obstacle avoidance types, optimizing the obstacle avoidance path templates and parameter matching rules in the obstacle avoidance strategy library, and incrementally updating the obstacle space occupancy information in the full-element digital twin model; The obstacle avoidance strategy library includes a process standard template library and an interference fault diagnosis template library. It simulates and extrapolates the motion trajectory of welding equipment and the spatial positional relationship between the welding torch sweeping space and obstacles under various working conditions. It collects spatial interaction data of each simulation node and simultaneously simulates the welding temperature field and stress field distribution under the corresponding working conditions. It analyzes the correlation between trajectory interference, component deformation and weld defects, generates standardized interference-free welding path samples and interference fault diagnosis reports, and stores them in the process standard template library and the interference fault diagnosis template library, respectively.

[0015] A further technical solution of the present invention is: the intelligent obstacle avoidance method for gate welding further includes an obstacle avoidance priority arbitration step: when the welding quality maintenance requirement conflicts with the obstacle avoidance safety requirement, arbitration is carried out based on a preset priority rule, the priority rule being the safety priority principle, that is, ensuring the safety of the welding equipment and gate components is the primary goal, and maximizing the stability of welding quality under the premise that safety is guaranteed; when multiple obstacles exist at the same time and the obstacle avoidance strategies conflict with each other, priority is ranked based on the degree of danger of each obstacle, and the obstacle with the highest degree of danger is avoided first.

[0016] A further technical solution of the present invention is: the welding equipment is a nine-axis linkage welding robot, which includes a robot body and two to seven additional motion axes; when generating the collision-free welding path in step S2, the digital twin model is imported into the robot joint kinematics constraint model, and the travel limits and rotation limits of each joint and additional axis are checked simultaneously to generate the optimal obstacle avoidance welding path that simultaneously satisfies the axis motion limit constraints and obstacle space avoidance constraints.

[0017] Another objective of this invention is to provide an intelligent obstacle avoidance system for gate welding based on digital twin technology, comprising: The model building module is used to collect three-dimensional design data of gate components, establish a full-element digital twin model containing geometric structure information, material property information, assembly relationship information and welding process parameter information, and build a virtual welding equipment model and a virtual working environment model in the simulation platform based on the kinematic parameters of the welding equipment. The path planning and pre-simulation module is used to automatically extract weld feature information based on the full-element digital twin model and generate an initial welding path and initial welding torch posture. In the simulation platform, it drives the virtual welding equipment model to move along the initial welding path to perform collision avoidance simulation analysis, calculates the spatial distance between each component in real time and determines the collision risk, and generates a collision-free welding path and corresponding obstacle avoidance process scheme through iterative optimization. The multi-source sensor perception module includes a vision sensor, a distance sensor, and a force sensor installed on the welding equipment. It is used to collect perception data of the welding operation area in real time during the welding process, identify static obstacle information and dynamic obstacle information based on the perception data, and calculate the relative pose relationship between the welding torch and each obstacle in real time. The dynamic obstacle avoidance and replanning module is used to map real-time sensing data to the full-element digital twin model after coordinate transformation, so that the pose of the virtual welding equipment is synchronized with the actual pose of the physical welding equipment. When an obstacle is detected to enter the preset safe distance range or a collision is predicted to occur, an obstacle avoidance adjustment strategy is generated to dynamically adjust the welding torch movement trajectory and / or welding parameters, control the welding equipment to perform obstacle avoidance actions, and automatically return to the planned position of the preset welding path after obstacle avoidance is completed. The closed-loop feedback and update module is used to feed back obstacle avoidance data, path adjustment data and welding quality inspection data generated during the actual welding process to the full-element digital twin model, and update the model parameters and obstacle avoidance strategy library through machine learning algorithms.

[0018] A further technical solution of the present invention is as follows: In the multi-source sensor perception module: the vision sensor includes a 3D structured light vision sensor and a laser contour scanning sensor, which are used to collect three-dimensional point cloud data of the welding operation area and two-dimensional cross-sectional data of the weld contour, respectively; the distance sensor includes an ultrasonic sensor and a laser ranging sensor, with a measurement accuracy of not less than ±0.5mm; the force sensor is a six-dimensional force / torque sensor, which is set at the connection flange between the welding torch and the welding robot body, with a sampling frequency of not less than 1000Hz; in the dynamic obstacle avoidance and replanning module, the virtual and actual pose synchronization frequency is not less than 10Hz, and the pose synchronization delay is not more than 50ms; the obstacle avoidance adjustment strategy includes one or more combinations of deceleration strategy, stopping strategy, offset strategy and attitude adjustment strategy, and the priority order of the obstacle avoidance adjustment strategy is: the combination of deceleration strategy and attitude adjustment strategy is preferred, followed by offset strategy, and finally stopping strategy; The system also includes an obstacle avoidance priority arbitration module, which arbitrates the conflict between welding quality maintenance requirements and obstacle avoidance safety requirements based on preset priority rules. The priority rules prioritize the safety of welding equipment and gate components. When multiple obstacles exist simultaneously and obstacle avoidance strategies conflict with each other, the system prioritizes the obstacles based on their degree of danger, avoiding the obstacle with the highest degree of danger first. The closed-loop feedback and update module uses a convolutional neural network to perform feature extraction and cluster analysis on multiple obstacle avoidance data, identifies high-frequency obstacle avoidance areas and high-frequency obstacle avoidance types, and optimizes the obstacle avoidance path templates and parameter matching rules in the obstacle avoidance strategy library. The dynamic obstacle avoidance and replanning module has a built-in interference warning unit, which is used to divide the interference warning level into three levels based on the predicted minimum distance between the welding torch and the obstacle: Level 1 is slight deviation, Level 2 is moderate interference risk and Level 3 is severe interference risk, and outputs deviation record, trajectory fine-tuning and emergency stop command respectively. The system also includes an XR visualization rendering unit, which is used to render the sweep space three-dimensional envelope of the moving parts of the welding equipment, the obstacle three-dimensional model, the obstacle avoidance path and the spatial safety gap value in real time, and push the rendered image to the on-site monitoring terminal in real time. The path planning and pre-simulation module has a built-in kinematic constraint verification unit, which is used to import the stroke limit parameters of each joint and additional axis of the welding equipment, and output the final interference-free obstacle avoidance path after verifying the axis motion limit constraints.

[0019] The beneficial effects of this invention are as follows: This application constructs an integrated digital twin model of the gate, welding equipment, and working environment, and completes offline collision avoidance simulation before the actual welding operation is carried out. Combined with the hierarchical bounding box collision detection algorithm, the risk of interference from static obstacles is identified in advance. At the same time, the kinematic limit verification of the nine-axis linkage robot is superimposed. The output collision-free welding path satisfies both spatial obstacle avoidance and the robot's own motion constraints, avoiding the problem that the simulation is feasible but the physical equipment cannot execute it, thus reducing the probability of collision from the source.

[0020] Employing a multi-source sensing fusion architecture, 3D structured light vision, laser profilometry, laser ranging, ultrasonic sensors, and six-dimensional force and torque sensors complement each other. This enables the identification of both static obstacles such as tooling fixtures and dynamic obstacles such as gate welding thermal deformation profiles, on-site personnel, and moving tooling. By relying on point cloud differential comparison, dynamic obstacles are automatically extracted, and force sensors serve as a contact fallback sensor, significantly improving the reliability of perception in complex gate welding conditions.

[0021] A high-frequency, low-latency virtual-real synchronous mapping link is established, and the future trajectory of the welding torch is predicted by combining the Kalman filter algorithm to achieve predictive obstacle avoidance, which is different from the traditional solution of passive handling after collision and contact. Three levels of interference warning are divided, multiple sets of combinable obstacle avoidance strategies are set and their execution priorities are clearly defined, and the use of methods with less interference to the weld seam, such as attitude adjustment, is given priority, taking into account both equipment hardware safety and welding continuity.

[0022] The obstacle avoidance action is synchronized with the welding process parameters. The current and voltage are dynamically adjusted according to the relative deviation between the welding torch and the weld to maintain stable heat input. After the obstacle avoidance is completed, a curvature-continuous interpolation transition path is used to smoothly return to the original welding trajectory, suppressing welding torch vibration and reducing the risk of weld defects caused by the obstacle avoidance action.

[0023] A complete closed-loop system of "modeling-planning-perception-obstacle avoidance-feedback" is established. All data of the obstacle avoidance process and weld quality inspection data are fed back into the digital twin model. Obstacle avoidance path templates are accumulated through machine learning clustering analysis, and obstacle occupancy information and obstacle avoidance strategy library are updated incrementally. The strategy library also stores standard non-interference samples and interference fault diagnosis reports, and is linked to welding temperature field and stress field simulations. This enables the correlation analysis between collision interference and welding quality defects. Subsequent gate welding tasks of the same type can directly reuse historical experience, reduce the workload of path debugging, and improve production efficiency.

[0024] Configure obstacle avoidance priority arbitration logic, clarify the underlying principle of safety priority, handle scenarios of conflict between welding quality and safety, and conflict between multiple obstacle strategies, and ensure the integrity of system decision-making logic under complex working conditions; add an XR visualization rendering unit to push the sweep envelope of virtual space and safety gaps to the on-site monitoring terminal, so that operation and maintenance personnel can intuitively monitor interference risks and improve human-machine collaboration capabilities. Attached Figure Description

[0025] Figure 1 This is a flowchart of the intelligent obstacle avoidance method for gate welding based on digital twin technology provided in an embodiment of the present invention. Detailed Implementation

[0026] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0027] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, 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, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0028] like Figure 1 The flowchart shown is a process for the intelligent obstacle avoidance method for gate welding based on digital twin technology provided by this invention. This intelligent obstacle avoidance method for gate welding relies on 3D modeling software, a finite element simulation platform, a nine-axis linkage welding robot, and a multi-source sensor acquisition system to realize gate welding operation path planning and obstacle avoidance; its details are as follows: Step S1: Collect 3D design data of the gate components and establish a full-element digital twin model containing geometric structure information, material property information, assembly relationship information, and welding process parameter information. This model is used to provide accurate digital benchmarks and data support for subsequent welding path planning, collision avoidance simulation analysis, and dynamic obstacle avoidance decision-making. Based on the kinematic parameters of the welding equipment, construct a virtual welding equipment model and a virtual working environment model that correspond one-to-one with the physical welding equipment in the simulation platform. The virtual welding equipment model includes at least a welding robot body model, an additional axis model, and a welding torch model. The construction of the virtual model is used to reproduce the motion characteristics and environmental constraints of the physical welding system in digital space, providing a high-fidelity simulation environment for offline path planning and collision pre-simulation.

[0029] A digital mirror image of the physical gate components, welding equipment, and operating environment is established in a virtual simulation space. This not only stores the geometric shape of the gate but also integrates multi-dimensional information on materials, assembly, and processes. The virtual welding equipment completely replicates the kinematic characteristics of the real equipment, providing a reliable simulation carrier for subsequent offline collision avoidance simulation. This allows the entire welding process to be reproduced in the virtual space before the physical entity has even started welding operations, thus identifying potential collision risks in advance.

[0030] Preferably, the full-element digital twin model is based on 3D modeling software to establish a modular, parametric 3D geometric model of the gate components. Geometric information includes weld groove dimensions, plate thickness, spatial geometric contour, and assembly gaps between components; used to accurately describe the spatial morphology and weld characteristics of the welding operation. Material property information includes the thermal conductivity, specific heat capacity, yield strength, and weldability parameters of commonly used gate steel; used to support temperature and stress field simulation analysis during the welding process. Assembly relationship information includes the connection methods, fit tolerances, and assembly sequence between components; used to assess the risk of spatial interference between components. Welding process parameter information includes the welding current range, welding voltage range, welding speed range, and welding torch angle range corresponding to different weld types; used to provide a data foundation for generating process plans.

[0031] The virtual work environment model includes the 3D spatial occupancy information of fixed obstacles around the gate components, the spatial occupancy information of temporary facilities, and the workspace boundary information of the welding equipment; it is used to construct a complete virtual work scene to support collision detection. The kinematic model of the welding robot body in the virtual welding equipment model is established using the Denavit-Hartenberg parametric method, and the additional axis model contains the kinematic parameters of seven to nine additional axes; it is used to accurately describe the kinematic characteristics of multi-axis linkage welding equipment. The mapping accuracy between the virtual welding equipment model and the physical welding equipment is no less than ±0.5mm; this ensures that the virtual simulation results have sufficient guiding significance for physical execution. The welding equipment is a nine-axis linkage welding robot, which includes the robot body and two to seven additional motion axes; the nine-axis linkage robot has multiple additional axes, and a purely spatial geometric, collision-free path may exceed the joint and additional axis rotation strokes.

[0032] The Denavit-Hartenberg parametric method fully describes the coordinate transformation relationship of each link in the robot, ensuring that the motion trajectory of the virtual robot is highly consistent with the motion height of the physical robot; the mapping accuracy index ensures that the collision distance calculated in the virtual simulation space can directly guide the physical equipment, avoiding excessive deviation between simulation and reality that would cause the simulation obstacle avoidance to fail.

[0033] Step S2 involves automatically extracting weld feature information based on a full-element digital twin model to generate an initial welding path and initial welding torch posture. This operation automatically converts weld information in the design model into an executable welding path, reducing the workload and errors of manual teaching. In the simulation platform, the virtual welding equipment model is driven to move along the initial welding path, and collision avoidance simulation analysis is performed. The spatial distances between each moving part of the welding equipment and the gate components and surrounding obstacles are calculated in real time. When any spatial distance is less than a preset safety threshold, a collision risk is identified, triggering an automatic adjustment strategy. The automatic adjustment strategy includes at least one of path offset adjustment, posture rotation adjustment, and welding sequence adjustment. Through iterative optimization, a collision-free welding path and corresponding obstacle avoidance process are generated. This pre-simulation operation is used to identify potential collision risks and generate an optimized path before welding execution, reducing the probability of collisions during the welding process from the source.

[0034] Before the physical equipment performs welding, the entire path is simulated in a virtual simulation environment to identify interference risks of the robot body, additional axes, welding gun, components, and tooling under the theoretical path in advance. The path is iteratively corrected through various adjustment methods to avoid most static obstacle collisions from the source. The output obstacle avoidance process solution not only includes the spatial trajectory, but also the welding gun posture and welding sequence, providing a safe and feasible initial process input for on-site physical welding.

[0035] Preferably, the collision avoidance simulation analysis employs a hierarchical bounding box collision detection algorithm. This algorithm performs hierarchical intersection tests on the bounding boxes of each moving component of the welding equipment and obstacles, calculating the minimum distance between each component. This algorithm improves the computational efficiency of collision detection while ensuring detection accuracy, meeting the real-time simulation requirements of large-scale complex scenarios. When the minimum distance is less than a preset safety threshold, the collision area is marked and the welding equipment's pose parameters at the moment of collision are recorded; this provides precise obstacle avoidance constraints for path optimization. The preset safety threshold ranges from 5mm to 20mm, with different safety thresholds set for different types of obstacles; this achieves a reasonable balance between safety and welding accessibility.

[0036] Path offset adjustment includes: maintaining the welding torch posture unchanged, and correcting the path coordinates in planes perpendicular and parallel to the welding direction, with the deviation after path correction controlled within ±0.5mm; used to achieve obstacle avoidance with minimal path alteration. Attitude rotation adjustment includes: maintaining the spatial coordinates of the welding path unchanged, and adjusting the pitch, yaw, and roll angles of the welding torch; optimizing the welding torch posture through additional axis linkage to avoid obstacles; the angle deviation between the welding torch axis and the weld centerline after adjustment ≤±1°; used to achieve obstacle avoidance by changing the welding torch direction rather than the path position, suitable for confined space scenarios. Welding sequence adjustment involves: adjusting the welding sequence of weld segments with collision risks, and adjusting high-risk areas to an unobstructed state for welding, used to avoid collision risks from a time perspective.

[0037] The hierarchical bounding box collision detection algorithm first performs coarse bounding box intersection screening, and then performs fine geometric detection, balancing collision detection speed and calculation accuracy; the differentiated safety threshold is adapted to the danger level of different obstacles in tooling fixtures and gate body components; the attitude rotation adjustment prioritizes ensuring that the weld trajectory point remains unchanged, only changing the welding torch orientation, minimizing the negative impact of path modification on weld formation quality.

[0038] For collision-free welding paths, the digital twin model imports the robot's joint kinematic constraint model to simultaneously verify the travel limits and angular limits of each joint and additional axis, generating an optimal obstacle-avoidance welding path that simultaneously satisfies axis motion limit constraints and obstacle space avoidance constraints. This operation ensures that the generated obstacle-avoidance path is kinematically feasible, preventing execution failure due to exceeding joint limits. Simultaneous motion axis limit verification during collision avoidance simulation ensures that the output path is both spatially interference-free and satisfies the robot's own motion capability constraints, guaranteeing that the simulated path can be actually executed by the physical robot.

[0039] Step S3: During the welding process, visual sensors, distance sensors, and force sensors installed on the welding equipment collect real-time perception data of the welding operation area. The fusion deployment of multiple sensors is used to comprehensively acquire multi-dimensional information about the welding operation environment, overcoming the limitations of a single sensor in terms of perception range, accuracy, and anti-interference capability. Based on the perception data, static and dynamic obstacle information is identified, and the relative pose relationship between the welding torch and each obstacle is calculated in real time. This operation is used to continuously perceive environmental changes during the welding process, providing real-time and accurate status input for dynamic obstacle avoidance decisions.

[0040] Preferably, the vision sensors include a 3D structured light vision sensor and a laser contour scanning sensor. The 3D structured light vision sensor is used to acquire three-dimensional point cloud data of the welding operation area to obtain three-dimensional spatial morphological information of obstacles and workpieces. The laser contour scanning sensor is used to acquire two-dimensional cross-sectional data of the weld contour, enabling precise positioning and tracking of the weld. The distance sensors include an ultrasonic sensor and a laser rangefinder, used for mid-to-long-range obstacle detection and close-range precise ranging, respectively, with a measurement accuracy of not less than ±0.5mm; they are used to achieve complementary obstacle perception within different distance ranges. The force sensor is a six-dimensional force / torque sensor, located at the connection flange between the welding torch and the welding robot body, used to acquire three-dimensional force and three-dimensional torque data when the welding torch contacts an obstacle in real time, with a sampling frequency of not less than 1000Hz; it is used to provide tactile-level collision detection and force feedback protection when vision and distance perception fail.

[0041] Static obstacle information includes the spatial location and geometric dimensions of the welding equipment mounting base, gate support fixtures, clamps, and surrounding fixed equipment; used to establish a static environment model to support path planning. Dynamic obstacle information includes dynamic contour changes caused by thermal deformation of gate components during welding, real-time location information of personnel moving around the welding equipment, and real-time location information of mobile devices; used to capture dynamic environmental changes during welding. The method for identifying dynamic obstacle information specifically involves: constructing a real-time 3D point cloud model of the welding operation area based on real-time 3D point cloud data collected by a visual sensor; performing point cloud registration and differential comparison between the real-time 3D point cloud model and the full-element digital twin model; extracting and clustering the difference regions to identify the spatial location and contour information of each dynamic obstacle. This method is used to accurately separate dynamic obstacles from the static background, avoiding misjudging inherent structures in the design model as obstacles.

[0042] Point cloud registration and differential comparison can compare the actual contours on site with the design model, and isolate areas with shape differences caused by thermal deformation and foreign objects, thus completing the automatic identification of dynamic obstacles; the six-dimensional force and torque sensor collects force signals at high frequency, serving as the last line of defense for hardware contact in extreme working conditions where visual ranging fails.

[0043] Step S4 involves mapping the real-time sensing data to a full-element digital twin model after coordinate transformation, ensuring that the pose of the virtual welding equipment is synchronized with the actual pose of the physical welding equipment. This virtual-real synchronization operation ensures that the digital twin model can reflect the real state of the physical world in real time, providing an accurate model basis for obstacle avoidance decisions. Based on the relative pose relationship, when an obstacle is detected entering the preset safe distance range or a collision is predicted, the digital twin model generates an obstacle avoidance adjustment strategy, dynamically adjusting the welding torch trajectory and / or dynamically adjusting welding parameters to control the welding equipment to perform obstacle avoidance actions. This dynamic obstacle avoidance operation is used to respond in real time to sudden obstacles or environmental changes during the welding process, maximizing the continuity of the welding operation while ensuring safety. After obstacle avoidance is completed, the system automatically returns to the planned position of the preset welding path. This return operation guides the welding torch to smoothly return to the original welding path after obstacle avoidance, ensuring the consistency of weld quality between the obstacle avoidance section and the non-obstacle avoidance section.

[0044] Physical field sensor data is sent to the digital twin virtual space in real time. The virtual device moves synchronously with the physical device, and collision trend is predicted within the virtual model before control commands are sent to the physical welding robot. This eliminates the need to wait for a collision to occur before taking action, enabling predictive obstacle avoidance. At the same time, the obstacle avoidance process is adapted to welding process parameters to avoid weld quality degradation caused by trajectory disturbances. After obstacle avoidance, a smooth transition trajectory is generated to return to the original weld path, avoiding welding defects caused by abrupt trajectory changes.

[0045] Preferably, mapping real-time perceived data to a full-element digital twin model after coordinate transformation includes: establishing a coordinate transformation matrix between the sensor coordinate system and the welding robot base coordinate system, transforming the perceived data collected by the vision sensor, distance sensor, and force sensor to a unified global coordinate system; this eliminates coordinate system differences between multiple sensors and achieves consistent data fusion. After demodulation and filtering, the data is mapped to the corresponding virtual device and virtual environment model in the simulation platform, ensuring that the pose of the virtual welding device is synchronized with the actual pose of the physical welding device. The virtual and actual pose synchronization frequency is not less than 10Hz, and the pose synchronization delay does not exceed 50ms; this ensures that the digital twin model has sufficient real-time response to changes in the physical world.

[0046] The method for predicting an impending collision is as follows: Based on the real-time relative pose between the welding torch and the obstacle, and the current speed and acceleration of the welding equipment, a Kalman filter algorithm is used to predict the trajectory of the welding torch within a preset time window. The predicted minimum distance between the predicted trajectory and the space occupied by the obstacle is calculated. When the predicted minimum distance is less than a preset safety threshold, a collision is determined to be imminent. The preset time window is 500ms to 2000ms, and the preset safety threshold is 10mm to 30mm. The safety threshold is dynamically adjusted according to the speed of the welding equipment; the higher the speed, the larger the safety threshold. This prediction method is used to anticipate risks before a collision occurs, allowing sufficient reaction time for obstacle avoidance decisions.

[0047] Three levels of interferometric early warning are defined based on the predicted minimum distance: Level 1 slight deviation, meaning the predicted minimum distance is greater than the first preset safety threshold, only records deviation data; used to maintain the continuity of welding operations when the risk is low.

[0048] Level 2 moderate interference risk, i.e. the predicted minimum distance is greater than 0 and less than or equal to the first preset safety threshold, outputs a trajectory fine-tuning instruction; used to adjust the path with minimal intervention when the risk is controllable.

[0049] Level 3 severe interference risk, meaning the predicted minimum distance is less than or equal to 0, outputs emergency stop and welding torch retraction commands; used to take mandatory safety measures when a collision is about to occur.

[0050] The three-level early warning system is used to achieve risk classification and control, avoid a "one-size-fits-all" obstacle avoidance response mode, and minimize interference with welding operations while ensuring safety.

[0051] The obstacle avoidance adjustment strategy is a combination of one or more of the following strategies: a) Deceleration strategy: Reduce the speed of the welding equipment to below the preset safe speed, with a deceleration time not exceeding 100ms; this is used to gain more reaction and decision-making time by reducing the speed of movement when a risk is detected.

[0052] b) Stopping strategy: Control the welding equipment to stop moving in an emergency, with a braking distance not exceeding 5mm; used to take mandatory stopping measures when a collision is unavoidable.

[0053] c) Offset strategy: Control the welding torch to offset along the normal direction of the obstacle surface by a preset offset amount to avoid the obstacle. The offset amount is dynamically calculated based on the obstacle's geometric dimensions; used to achieve obstacle avoidance by changing the path position.

[0054] d) Attitude adjustment strategy: By adjusting the pitch and yaw angles of the welding torch, the spatial orientation of the welding torch is changed, so that the welding torch avoids obstacles while maintaining welding operation capability; it is used to achieve obstacle avoidance by changing the attitude of the welding torch without changing the path position.

[0055] The priority order of obstacle avoidance adjustment strategies is as follows: the combination of deceleration strategy and attitude adjustment strategy is preferred; when attitude adjustment cannot meet the obstacle avoidance requirements, the offset strategy is activated; and when the offset strategy still cannot avoid the obstacle, the stopping strategy is activated. This priority order is used to select the obstacle avoidance method with the least impact on welding quality under the premise of ensuring safety.

[0056] Dynamic adjustment of welding parameters includes: during obstacle avoidance, dynamically adjusting the welding current and welding voltage based on the relative posture deviation between the welding torch and the weld to maintain a stable heat input and ensure consistent weld quality between the obstacle avoidance and non-obstacle avoidance sections; this operation compensates for the impact of welding torch posture changes caused by obstacle avoidance on weld quality. After obstacle avoidance is completed, the digital twin model generates a smooth transition path from the current position back to the preset welding path using a path interpolation algorithm, guiding the welding equipment along the transition path back to the preset welding path to ensure the smoothness of the return process and the stability of equipment operation. The planned position of the transition path is such that the curvature of the transition path is continuous and the rate of acceleration change does not exceed a preset comfort threshold.

[0057] Kalman filtering, combined with the equipment's motion state, predicts the future trajectory of the welding torch, enabling pre-judgment of collision risks. Synchronization frequency and delay indicators ensure that the virtual and real states are almost identical, avoiding obstacle avoidance failure caused by the virtual space calculation results lagging behind the physical movement. A three-level early warning and graded handling system matches different handling actions to avoid the overprotection problem of directly shutting down the machine upon detecting a risk. The strategy priority prioritizes attitude adjustments that do not deviate from the weld seam trajectory, maximizing the maintenance of welding continuity. A smooth transition path prevents welding torch jitter caused by sudden acceleration changes, protecting the weld seam formation quality.

[0058] Step S5 involves feeding back obstacle avoidance data, path adjustment data, and welding quality inspection data generated during the actual welding process to the full-element digital twin model, updating the model parameters and obstacle avoidance strategy library, and forming a closed-loop optimization system of "modeling-planning-perception-obstacle avoidance-feedback". This closed-loop feedback operation is used to transform the experience data accumulated during on-site execution into updatable knowledge for the model, enabling the system to have continuous learning and self-optimization capabilities, and continuously improving the accuracy of obstacle avoidance decisions and the rationality of process planning.

[0059] All obstacle conditions, obstacle avoidance processes, and corresponding weld quality results encountered in a single welding task are collected and stored. High-frequency interference areas and effective obstacle avoidance templates are mined through machine learning, and the obstacle information and obstacle avoidance strategy library of the twin model are incrementally updated. Subsequent similar gate welding tasks can directly reuse historical experience, reduce the workload of repeated simulation and debugging, and realize the system's obstacle avoidance capability to continuously improve with production data.

[0060] Obstacle avoidance data includes the timestamp of the obstacle avoidance trigger moment, the 3D coordinates of the obstacle avoidance trigger location, the sequence of path trajectory points during obstacle avoidance execution, the welding parameter adjustment record during obstacle avoidance, and the timestamp of the obstacle avoidance completion moment; used to completely record the spatiotemporal information and execution details of the obstacle avoidance event. Path adjustment data includes the deviation data and deviation correction amount between the initial welding path and the actual execution path after obstacle avoidance; used to quantify the degree of path deviation and guide subsequent path planning. Welding quality inspection data includes weld formation size inspection data and weld internal defect inspection data; used to evaluate the impact of obstacle avoidance on welding quality. Updating model parameters and the obstacle avoidance strategy library specifically includes: performing cluster analysis on multiple obstacle avoidance data using machine learning algorithms to identify high-frequency obstacle avoidance areas and high-frequency obstacle avoidance types, optimizing the obstacle avoidance path templates and parameter matching rules in the obstacle avoidance strategy library, and incrementally updating the obstacle space occupancy information in the full-element digital twin model; this operation is used to transform historical obstacle avoidance experience into reusable knowledge assets, continuously improving the system's obstacle avoidance capabilities and path planning efficiency.

[0061] The obstacle avoidance strategy library includes a process standard template library and an interference fault diagnosis template library. It simulates and extrapolates the motion trajectory of welding equipment and the spatial relationship between the welding torch sweeping space and obstacles under various working conditions. It collects spatial interaction data from each simulation node and simultaneously simulates the welding temperature field and stress field distribution under corresponding working conditions. It analyzes the correlation between trajectory interference, component deformation, and weld defects, generating standardized interference-free welding path samples and interference fault diagnosis reports, which are stored in the process standard template library and the interference fault diagnosis template library, respectively. This operation is used to establish a quantitative correlation model between welding path planning and welding quality, providing multi-dimensional evaluation basis for obstacle avoidance decisions.

[0062] The strategy library is divided into a process standard master library and an interference fault diagnosis master library. On the one hand, it stores feasible non-interference path samples, and on the other hand, it stores the fault diagnosis logic corresponding to interference conditions. At the same time, it is associated with temperature field and stress field simulation to establish the correspondence between collision interference, component deformation and weld defects. This not only realizes equipment collision prevention, but also helps to predict the welding quality hazards that interference may bring.

[0063] Step S6: When the requirements for maintaining welding quality conflict with the requirements for obstacle avoidance safety, arbitration is conducted based on preset priority rules. The priority rule is the principle of safety first, that is, ensuring the safety of welding equipment and gate components is the primary goal, and maximizing the stability of welding quality under the premise that safety is guaranteed. When multiple obstacles exist simultaneously and obstacle avoidance strategies conflict with each other, priority is given to avoiding the obstacle with the highest degree of danger based on the danger level of each obstacle. This arbitration step is used to provide a clear decision-making basis in multi-objective conflict scenarios and avoid mutual interference between obstacle avoidance strategies.

[0064] Another objective of this invention is to provide an intelligent obstacle avoidance system for gate welding based on digital twin technology, comprising: The model building module is used to collect three-dimensional design data of gate components, establish a full-element digital twin model containing geometric structure information, material property information, assembly relationship information and welding process parameter information, and build a virtual welding equipment model and a virtual working environment model in the simulation platform based on the kinematic parameters of the welding equipment. The path planning and pre-simulation module is used to automatically extract weld feature information based on the full-element digital twin model and generate an initial welding path and initial welding torch posture. In the simulation platform, it drives the virtual welding equipment model to move along the initial welding path to perform collision avoidance simulation analysis, calculates the spatial distance between each component in real time and determines the collision risk, and generates a collision-free welding path and corresponding obstacle avoidance process scheme through iterative optimization. The multi-source sensor perception module includes a vision sensor, a distance sensor, and a force sensor installed on the welding equipment. It is used to collect perception data of the welding operation area in real time during the welding process, identify static obstacle information and dynamic obstacle information based on the perception data, and calculate the relative pose relationship between the welding torch and each obstacle in real time. The dynamic obstacle avoidance and replanning module is used to map the real-time sensing data to the full-element digital twin model after coordinate transformation, so that the pose of the virtual welding equipment is synchronized with the actual pose of the physical welding equipment. When an obstacle is detected to enter the preset safe distance range or a collision is predicted to occur, an obstacle avoidance adjustment strategy is generated to dynamically adjust the welding torch movement trajectory and / or welding parameters, control the welding equipment to perform obstacle avoidance actions, and automatically return to the planned position of the preset welding path after obstacle avoidance is completed. The closed-loop feedback and update module is used to feed back obstacle avoidance data, path adjustment data and welding quality inspection data generated during the actual welding process to the full-element digital twin model, and update the model parameters and obstacle avoidance strategy library through machine learning algorithms.

[0065] Preferably, in the multi-source sensor perception module: the vision sensors include a 3D structured light vision sensor and a laser contour scanning sensor, which are used to collect three-dimensional point cloud data of the welding operation area and two-dimensional cross-sectional data of the weld contour, respectively; the distance sensors include an ultrasonic sensor and a laser ranging sensor, with a measurement accuracy of not less than ±0.5mm; the force sensor is a six-dimensional force / torque sensor, which is set at the connection flange between the welding torch and the welding robot body, with a sampling frequency of not less than 1000Hz; in the dynamic obstacle avoidance and replanning module, the virtual and actual pose synchronization frequency is not less than 10Hz, and the pose synchronization delay is not more than 50ms; the obstacle avoidance adjustment strategy includes one or more combinations of deceleration strategy, stopping strategy, offset strategy and attitude adjustment strategy, and the priority order of the obstacle avoidance adjustment strategy is: the combination of deceleration strategy and attitude adjustment strategy is preferred, followed by offset strategy, and finally stopping strategy; The system also includes an obstacle avoidance priority arbitration module, which arbitrates the conflict between welding quality maintenance requirements and obstacle avoidance safety requirements based on preset priority rules. The priority rules prioritize the safety of welding equipment and gate components. When multiple obstacles exist simultaneously and obstacle avoidance strategies conflict, the system prioritizes the obstacles based on their degree of danger, avoiding the obstacle with the highest degree of danger first. The closed-loop feedback and update module uses a convolutional neural network to perform feature extraction and cluster analysis on multiple obstacle avoidance data, identify high-frequency obstacle avoidance areas and high-frequency obstacle avoidance types, and optimize the obstacle avoidance path templates and parameter matching rules in the obstacle avoidance strategy library. The dynamic obstacle avoidance and replanning module has a built-in interference warning unit, which is used to divide the interference warning level into three levels based on the predicted minimum distance between the welding torch and the obstacle: Level 1 is slight deviation, Level 2 is moderate interference risk and Level 3 is severe interference risk, and the corresponding outputs are deviation record, trajectory fine-tuning and emergency stop command respectively. The system also includes an XR visualization rendering unit, which is used to render the sweep space three-dimensional envelope of the moving parts of the welding equipment, the obstacle three-dimensional model, the obstacle avoidance path and the spatial safety gap value in real time, and push the rendered screen to the on-site monitoring terminal in real time. The path planning and pre-simulation module has a built-in kinematic constraint verification unit, which is used to import the stroke limit parameters of each joint and additional axis of the welding equipment, and output the final interference-free obstacle avoidance path after verifying the axis motion limit constraints.

[0066] Example 1 1. System Hardware Configuration The hardware devices used in this embodiment include: Welding equipment: a nine-axis linkage welding robot, which includes the robot body and five additional motion axes (two to seven additional axes can be configured according to actual needs). Each joint and additional axis is equipped with a high-precision encoder and servo drive system to achieve precise movement of complex spatial trajectories.

[0067] 3D Design and Simulation Platform: Equipped with 3D modeling software and a digital twin simulation system, used to create full-element digital twin models of gate components and a virtual operating environment.

[0068] Multi-source sensor system: including 3D structured light vision sensor (resolution not less than 1920×1080, depth accuracy not less than ±0.1mm), laser contour scanning sensor, ultrasonic sensor (detection distance 0.1m~10m), laser rangefinder sensor (accuracy ±0.5mm) and six-dimensional force / torque sensor (range determined according to the contact force between the welding torch and the workpiece, sampling frequency 1000Hz).

[0069] Computing and Control Platform: A digital twin real-time computing server used to run digital twin models, collision detection algorithms, path planning algorithms, and obstacle avoidance decision-making algorithms, and communicates in real time with the welding robot control system via industrial Ethernet.

[0070] Quality inspection equipment: laser scanning detector (scanning accuracy ±0.1mm) and ultrasonic flaw detector, used for weld formation size inspection and internal defect detection after welding is completed.

[0071] 2. Execution process of intelligent obstacle avoidance method for gate welding Step S1: Establish a full-element digital twin model of a large-scale hydraulic arc gate based on 3D modeling software. The gate leaf plate is 20mm to 32mm thick, the side beam adopts a box-type welded structure, the support arm has an I-shaped cross section, and the total length of the weld is about 56m, including various joint forms such as fillet weld, vertical weld, overhead weld, and multi-layer multi-pass weld.

[0072] Geometric structure information modeling: Using a modular and parametric design method, the gate is divided into component modules such as gate leaf, side beam, support arm, and connecting plate. Three-dimensional geometric models of each component are established to clarify key geometric parameters such as weld bevel angle (45°~60°), plate thickness (20mm~32mm), assembly gap (0~2mm), and spatial assembly relationship between components.

[0073] Material property information embedding: embedding the thermophysical and mechanical properties of commonly used gate steels Q355B and Q390E, including: thermal conductivity (approximately 45 W / (m·K) for Q355B), specific heat capacity (approximately 500 J / (kg·K)), yield strength (355 MPa for Q355B and 390 MPa for Q390E), coefficient of linear expansion, elastic modulus, and weldability parameters (carbon equivalent, cold crack sensitivity coefficient, etc.).

[0074] Welding process parameter information configuration: A welding process window database is established by combining a large amount of welding process test data and national standards. For different types of welds (flat fillet welds, vertical welds, overhead welds, multi-layer multi-pass welds) and different plate thicknesses (20mm~32mm), corresponding welding current ranges (200~380A), welding voltage ranges (24~34V), welding speed ranges (3~9mm / s), welding torch angle ranges, and interpass temperature control requirements are configured respectively.

[0075] Virtual welding equipment model construction: Based on the kinematic parameters of the nine-axis linkage welding robot (DH parameters of each joint, motion range and transmission ratio of additional axes, etc.), a virtual welding equipment model corresponding one-to-one with the physical equipment is constructed in the simulation platform. The kinematic model of the welding robot body is established using the Denavit-Hartenberg parametric method, and the additional axis model includes the kinematic parameters of 5 additional axes (which can be expanded to seven to nine axes as needed). The mapping accuracy between the virtual welding equipment model and the physical welding equipment is controlled within ±0.5mm.

[0076] Virtual work environment model construction: In the simulation platform, construct the three-dimensional spatial occupancy information of fixed obstacles around the gate components (such as welding equipment installation base, operating platform, column, etc.), the spatial occupancy information of temporary facilities (such as welding fixtures, jigs, support frames, etc.), and the workspace boundary information of the welding equipment to form a complete virtual work scene.

[0077] Step S2, Automatic Extraction of Weld Feature Information: Based on the constructed full-element digital twin model, the weld feature information of each component of the gate is automatically extracted, including the spatial coordinates of the weld, the weld length (a total of 24 main welds were identified, with a total length of about 56m), the weld thickness (ranging from 8mm to 32mm), the weld type (flat fillet weld, vertical weld, overhead weld, multi-layer multi-pass weld) and the joint type (butt joint, T-joint, corner joint, etc.).

[0078] Initial welding path generation: Based on weld feature information and the predetermined welding sequence ("inside first, then outside, symmetrical welding, segmented skip welding"), the initial welding path and initial welding torch posture are generated. The initial welding parameters are set as follows: welding current 300A, welding voltage 28V, welding speed 5mm / s, and welding torch tilt angle 15°.

[0079] Collision avoidance simulation analysis: In the simulation platform, the virtual welding equipment model is driven to move along the initial welding path. A hierarchical bounding box collision detection algorithm is used to perform hierarchical intersection tests on the bounding boxes of each moving part of the welding equipment (robot body, joints, additional axes, welding torch, etc.) and obstacles. The minimum distance between each part is calculated in real time, and the preset safety thresholds are set to 10mm (for fixed obstacles) and 15mm (for dynamic obstacles such as personnel).

[0080] Collision Risk Identification and Path Optimization: Simulation analysis revealed that at the connection between the side beam and the support arm (welds 8 and 12), the minimum distance between the welding torch and the vertical support rod of the supporting fixture near the weld end was only 4mm, less than the preset safety threshold of 10mm, indicating a collision risk. After recording the pose parameters of the collision area, an automatic adjustment strategy was triggered. First, we tried adjusting the attitude rotation—keeping the spatial coordinates of the welding path unchanged, we adjusted the pitch angle of the welding torch from 15° to 22° and the deflection angle from 0° to 8°, and optimized the attitude of the welding torch through the additional axis linkage. After adjustment, the angle deviation between the welding torch axis and the weld centerline was 0.8° (≤±1° requirement). The collision detection was re-performed, showing that the minimum distance increased to 12mm, meeting the safety requirements.

[0081] After iterative optimization, a collision-free welding path and corresponding obstacle avoidance process scheme are generated (welding of welds No. 8 and No. 12 is performed using the adjusted welding gun posture).

[0082] Step S3: During the welding process, the multi-source sensor system is activated for real-time sensing. Visual perception: The 3D structured light vision sensor acquires three-dimensional point cloud data of the welding operation area at a frequency of 10Hz (point cloud density not less than 100 points / cm²), and the laser contour scanning sensor acquires two-dimensional cross-sectional data of the weld contour at a frequency of 50Hz (lateral resolution not less than 0.02mm).

[0083] Distance sensing: Ultrasonic sensors detect obstacles at medium to long distances (0.5m to 5m) at a frequency of 20Hz, while laser rangefinders accurately measure distances at close ranges (0.05m to 0.5m) at a frequency of 100Hz.

[0084] Force perception: The six-dimensional force / torque sensor collects three-dimensional force and three-dimensional torque data in real time when the welding torch comes into contact with the obstacle at a sampling frequency of 1000Hz.

[0085] Static obstacle recognition: Based on laser scanning, a static obstacle database (including welding equipment mounting base, gate support fixtures, jigs, etc.) is pre-established. During the welding process, the spatial position and geometric dimensions of static obstacles are identified through point cloud registration and differential comparison.

[0086] Dynamic obstacle recognition: The real-time 3D point cloud model is registered and differentially compared with the full-element digital twin model (registration accuracy not less than ±0.5mm). The difference regions are extracted and segmented using the Euclidean distance clustering algorithm. During a welding process, the system successfully identified an operator entering the welding operation area (the point cloud cluster volume is approximately 0.6m × 0.4m × 1.7m, located about 800mm to the right of the welding torch's forward direction), as well as the change in the gate component's contour due to thermal deformation (the outer side beam shifted outward by about 1.2mm due to thermal expansion).

[0087] Real-time calculation of the relative positional relationship between the welding torch and various obstacles: the minimum distance between the welding torch and the operator is about 750mm, and the minimum distance between the welding torch and the outer edge of the heat-deformed side beam is about 18mm.

[0088] Step S4, Data Mapping and Virtual-Real Synchronization: Establish a coordinate transformation matrix (including a 3×3 rotation matrix and a 3×1 translation vector) between the sensor coordinate system and the welding robot's base coordinate system. Transform the sensor data collected by the vision sensor, distance sensor, and force sensor to a unified global coordinate system. After demodulation and filtering, map the data to the virtual device and virtual environment model in the simulation platform at a synchronization frequency of 20Hz. The pose synchronization delay between the virtual and actual models is controlled within 30ms.

[0089] Collision prediction: Based on the real-time relative pose between the welding torch and obstacles, and the current velocity (5mm / s) and acceleration (50mm / s²) of the welding equipment, a Kalman filter algorithm is used to predict the trajectory of the welding torch within a 1000ms time window. The predicted minimum distance between the predicted trajectory and the operator's spatial position is calculated, and the result is 720mm (far greater than the 30mm safety threshold, thus deemed safe); the predicted minimum distance between the predicted trajectory and the outer edge of the heat-deformed side beam is 15mm (less than the 30mm safety threshold, thus deemed an impending collision).

[0090] Level 3 Early Warning and Obstacle Avoidance Decision: Based on the predicted minimum distance of 15mm, a Level 2 moderate interference risk warning is triggered (0 < predicted minimum distance ≤ 30mm), and the digital twin outputs a slow trajectory fine-tuning instruction.

[0091] Obstacle avoidance strategy execution: Following priority, a combination of deceleration and attitude adjustment strategies is first employed—reducing the welding speed from 5 mm / s to 3.5 mm / s (deceleration time 80 ms), while simultaneously adjusting the welding torch's pitch angle from 22° to 18° and its deflection angle from 8° to 5°, restoring the actual minimum distance between the welding torch and the outer edge of the heat-deformed side beam to 22 mm. If attitude adjustment fails to meet obstacle avoidance requirements (e.g., if the heat deformation of subsequent welding sections further increases to 2.5 mm), an offset strategy is activated—controlling the welding torch to offset outwards by 2 mm along the normal direction of the obstacle surface, maintaining the minimum distance above 25 mm.

[0092] Dynamic adjustment of welding parameters: During the obstacle avoidance process, based on the relative positional deviation between the welding torch and the weld (the welding torch offset is about 1.5mm, and the pitch angle changes by 4°), the welding current is adjusted from 300A to 285A and the welding voltage is adjusted from 28V to 27.5V to maintain a stable heat input and ensure the consistency of weld quality between the obstacle avoidance section and the non-obstacle avoidance section.

[0093] Path Return: After obstacle avoidance is completed (approximately 15 seconds after thermal deformation stabilizes), the digital twin model generates a transition path that smoothly returns from the current position to the preset welding path using a cubic spline path interpolation algorithm. This guides the welding equipment back to its original planned position along the transition path. The curvature of the transition path is continuous, and the rate of change of acceleration is controlled within 200 mm / s³.

[0094] Step S5, Data Collection and Recording: Completely record all relevant data for this obstacle avoidance event. Obstacle avoidance data: timestamp of obstacle avoidance triggering time (187 seconds after welding started), three-dimensional coordinates of obstacle avoidance triggering position (X=1245.3mm, Y=876.2mm, Z=342.1mm), path trajectory point sequence during obstacle avoidance execution (a total of 236 trajectory points recorded), welding parameter adjustment records during obstacle avoidance process (current and voltage adjustment curves), and timestamp of obstacle avoidance completion time (203 seconds after welding started).

[0095] Path adjustment data: The maximum deviation between the initial welding path and the actual execution path after obstacle avoidance is 2.3mm (located at the outer corner of the side beam), and the deviation correction amount is 1.8mm.

[0096] Welding quality inspection data: After welding, laser scanning inspection showed that the weld reinforcement height of the obstacle avoidance section was 2.0mm~3.5mm (design allowable range 1.5mm~4.0mm), the weld width was 8.5mm~10.5mm (design allowable range 8mm~12mm), and the undercut depth was ≤0.3mm; ultrasonic testing did not find internal defects such as cracks, incomplete penetration, or porosity, and there was no significant difference in weld quality between the obstacle avoidance section and the non-obstacle avoidance section.

[0097] Model Update: The above data was fed back into the full-element digital twin model. Convolutional neural networks were used to extract features and perform cluster analysis on multiple obstacle avoidance data points, identifying the outer region of the side beam as a high-frequency obstacle avoidance area (this area experienced 7 obstacle avoidance events in the last 20 welding operations), with the main obstacle avoidance type being "dynamic obstacle interference caused by thermal deformation." Based on this analysis, the obstacle avoidance path template for the corresponding area in the obstacle avoidance strategy library was optimized—the preset safety threshold for the outer region of the side beam was adjusted from 10mm to 15mm, and a 2mm obstacle avoidance margin was reserved for this area during the path planning stage. Simultaneously, the obstacle space occupancy information in the full-element digital twin model was incrementally updated, incorporating the measured data of this thermal deformation (maximum deformation of 2.5mm, deformation direction outwards) into the model parameters.

[0098] Through continuous iterative optimization (a total of 12 welding operations with closed-loop feedback), the pre-simulation deviation of this type of weld has been reduced from the initial 8.5% to 3.2%, and the obstacle avoidance trigger frequency in the high-frequency obstacle avoidance area has been reduced from 35% to 12%, significantly improving the system's obstacle avoidance capability and welding efficiency.

[0099] In step S6, during a welding operation, the system simultaneously detects two obstacle risks: an operator is moving approximately 600mm to the right (predicted minimum distance 550mm, classified as a Level 1 minor deviation), and a gate component protrudes approximately 20mm directly in front due to thermal deformation (predicted minimum distance 12mm, classified as a Level 2 moderate interference risk). The obstacle avoidance priority arbitration module prioritizes obstacles based on their degree of danger, determining that the thermally deformed obstacle directly in front poses a higher risk (closer distance, higher probability of collision). Obstacle avoidance strategies (attitude adjustment + deceleration) are implemented for this obstacle first, while monitoring of the operator on the right is maintained (only deviation data is recorded). When the operator moves closer to within 30mm of the predicted minimum distance, the risk is upgraded to Level 2, and obstacle avoidance measures are implemented simultaneously.

[0100] After practical application and verification in a large-scale water conservancy gate manufacturing enterprise, the intelligent obstacle avoidance method and system for gate welding based on digital twins described in this invention has achieved significant results: The incidence of collision accidents during welding is reduced by 92% (from an average of 3.2 collisions per 100 welds in the traditional method to 0.25 collisions); the average welding interruption time due to obstacle avoidance is reduced by 75%; the weld quality consistency between obstacle-avoidance sections and non-obstacle-avoidance sections reaches over 95% (based on a comprehensive evaluation of multiple indicators such as weld forming dimensions and internal defects); the overall efficiency of welding operations is improved by approximately 28%; and the labor intensity and skill requirements of operators are significantly reduced.

[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A gate welding intelligent obstacle avoidance method based on digital twin technology, which utilizes 3D modeling software, a finite element simulation platform, a nine-axis linkage welding robot, and a multi-source sensor acquisition system to achieve gate welding operation path planning and obstacle avoidance; characterized in that, The intelligent obstacle avoidance method for gate welding includes the following steps: S1. Collect the three-dimensional design data of the gate components and establish a full-element digital twin model containing geometric structure information, material property information, assembly relationship information and welding process parameter information; based on the kinematic parameters of the welding equipment, construct a virtual welding equipment model and a virtual working environment model that correspond one-to-one with the physical welding equipment in the simulation platform. The virtual welding equipment model includes at least a welding robot body model, an additional axis model and a welding torch model. S2. Based on the full-element digital twin model, automatically extract weld feature information to generate the initial welding path and initial welding torch posture; drive the virtual welding equipment model to move along the initial welding path in the simulation platform, perform collision avoidance simulation analysis, calculate the spatial distance between each moving part of the welding equipment and the gate component and surrounding obstacles in real time, and determine that there is a collision risk when any spatial distance is less than the preset safety threshold, triggering an automatic adjustment strategy. The automatic adjustment strategy includes at least one of path offset adjustment, posture rotation adjustment and welding sequence adjustment, and generates a collision-free welding path and corresponding obstacle avoidance process scheme through iterative optimization. S3. During the welding process, visual sensors, distance sensors, and force sensors installed on the welding equipment collect real-time perception data of the welding operation area. Based on the perception data, static obstacle information and dynamic obstacle information are identified, and the relative pose relationship between the welding torch and each obstacle is calculated in real time. S4. The real-time sensing data is mapped to a full-element digital twin model after coordinate transformation, so that the pose of the virtual welding equipment is synchronized with the actual pose of the physical welding equipment. Based on the relative pose relationship, when an obstacle is detected to enter the preset safe distance range or a collision is predicted, the digital twin model generates an obstacle avoidance adjustment strategy, dynamically adjusts the welding torch trajectory and / or dynamically adjusts the welding parameters, controls the welding equipment to perform obstacle avoidance actions, and automatically returns to the planned position of the preset welding path after obstacle avoidance is completed. S5. Feed back obstacle avoidance data, path adjustment data and welding quality inspection data generated during the actual welding process to the full-element digital twin model, update the model parameters and obstacle avoidance strategy library, and form a closed-loop optimization system of "modeling-planning-perception-obstacle avoidance-feedback".

2. The intelligent obstacle avoidance method for gate welding based on digital twin technology according to claim 1, characterized in that, In step S1, the full-element digital twin model is established based on 3D modeling software to create a modular and parametric 3D geometric model of the gate components. The geometric structure information includes the weld groove size, plate thickness, spatial geometric contour, and assembly gap between components of the gate components; the material property information includes the thermal conductivity, specific heat capacity, yield strength, and weldability parameters of commonly used steels for gates; the assembly relationship information includes the connection method, fit tolerance, and assembly sequence between each component; and the welding process parameter information includes the welding current range, welding voltage range, welding speed range, and welding torch angle range corresponding to different weld types. The virtual working environment model includes three-dimensional spatial occupancy information of fixed obstacles around the gate component, spatial occupancy information of temporary facilities, and workspace boundary information of the welding equipment; the kinematic model of the welding robot body in the virtual welding equipment model is established using the Denavit-Hartenberg parametric method, the additional axis model includes kinematic parameters of seven to nine additional axes, and the mapping accuracy between the virtual welding equipment model and the physical welding equipment is not less than ±0.5mm.

3. The intelligent obstacle avoidance method for gate welding based on digital twin technology according to claim 1, characterized in that, In step S2, the collision avoidance simulation analysis adopts a hierarchical bounding box collision detection algorithm, which performs hierarchical intersection tests on the bounding box hierarchical structure of each moving part of the welding equipment and obstacles, and calculates the minimum distance between each part; when the minimum distance is less than a preset safety threshold, the collision area is marked and the welding equipment pose parameters at the moment of collision are recorded. The preset safety threshold is 5mm to 20mm, and different safety thresholds are set for different types of obstacles. The path offset adjustment includes: keeping the welding torch posture unchanged, and offsetting and correcting the path coordinates in planes perpendicular to and parallel to the welding forward direction, with the deviation after path correction controlled within ±0.5mm; the posture rotation adjustment includes: keeping the spatial coordinates of the welding path unchanged, adjusting the pitch angle, yaw angle, and roll angle of the welding torch, and optimizing the welding torch posture through additional axis linkage to avoid obstacles, with the angle deviation between the welding torch axis and the weld centerline ≤±1° after adjustment; the welding sequence adjustment includes: adjusting the welding sequence of weld segments with collision risks, and adjusting high-risk areas to a state of no obstacles for welding.

4. The intelligent obstacle avoidance method for gate welding based on digital twin technology according to claim 1, characterized in that, The visual sensors in step S3 include a 3D structured light visual sensor and a laser contour scanning sensor. The 3D structured light visual sensor is used to collect three-dimensional point cloud data of the welding operation area, and the laser contour scanning sensor is used to collect two-dimensional cross-sectional data of the weld contour. The distance sensors include an ultrasonic sensor and a laser ranging sensor, which are used for mid-to-long-range obstacle detection and close-range precise ranging, respectively, with a measurement accuracy of not less than ±0.5mm. The force sensor is a six-dimensional force / torque sensor, which is set at the connection flange between the welding torch and the welding robot body, and is used to collect three-dimensional force and three-dimensional torque data when the welding torch contacts the obstacle in real time, with a sampling frequency of not less than 1000Hz. The static obstacle information includes the spatial position and geometric dimensions of the welding equipment mounting base, gate support fixture, clamps, and surrounding fixed equipment; the dynamic obstacle information includes the dynamic contour changes caused by thermal deformation of the gate components during welding, the real-time position information of personnel moving around the welding equipment, and the real-time position information of mobile devices; the method for identifying the dynamic obstacle information is as follows: a real-time three-dimensional point cloud model of the welding operation area is constructed based on the real-time three-dimensional point cloud data collected by the visual sensor; the real-time three-dimensional point cloud model is registered and differentially compared with the full-element digital twin model; the difference regions are extracted and clustered to segment the difference regions, thereby identifying the spatial position and contour information of each dynamic obstacle.

5. The intelligent obstacle avoidance method for gate welding based on digital twin technology according to claim 1, characterized in that, The step S4, which involves mapping the real-time sensing data to the full-element digital twin model after coordinate transformation, includes: establishing a coordinate transformation matrix between the sensor coordinate system and the welding robot base coordinate system; transforming the sensing data collected by the vision sensor, distance sensor, and force sensor to a unified global coordinate system; and mapping the data to the corresponding virtual device and virtual environment model in the simulation platform after demodulation and filtering, so that the pose of the virtual welding device is synchronized with the actual pose of the physical welding device, with a virtual and actual pose synchronization frequency of not less than 10Hz and a pose synchronization delay of not more than 50ms. The method for predicting an impending collision is as follows: based on the real-time relative pose between the welding torch and the obstacle, and the current speed and acceleration of the welding equipment, a Kalman filter algorithm is used to predict the trajectory of the welding torch within a preset time window. The predicted minimum distance between the predicted trajectory and the space occupied by the obstacle is calculated. When the predicted minimum distance is less than a preset safety threshold, it is determined that a collision is imminent. The preset time window is 500ms to 2000ms, and the preset safety threshold is 10mm to 30mm. The safety threshold is dynamically adjusted according to the speed of the welding equipment; the higher the speed, the larger the safety threshold. In step S4, the three levels of interferometry early warning are determined based on the predicted minimum distance: Level 1 slight deviation, meaning the predicted minimum distance is greater than the first preset safety threshold, only the deviation data is recorded; Level 2 moderate interference risk, i.e. the predicted minimum distance is greater than 0 and less than or equal to the first preset safety threshold, output trajectory slow fine-tuning instruction; Level 3 severe interference risk, i.e. the predicted minimum distance is less than or equal to 0, outputs emergency stop and welding torch retraction commands; The obstacle avoidance adjustment strategy is a combination of one or more of the following strategies: a) Deceleration strategy: Reduce the speed of the welding equipment to below the preset safe speed, and the deceleration time shall not exceed 100ms; b) Stopping strategy: Control the welding equipment to stop its movement in an emergency, with a braking distance not exceeding 5mm; c) Offset strategy: Control the welding torch to offset by a preset amount along the normal direction of the obstacle surface to avoid the obstacle. The offset amount is dynamically calculated based on the geometric dimensions of the obstacle. d) Attitude adjustment strategy: By adjusting the pitch and yaw angles of the welding torch, the spatial orientation of the welding torch can be changed, so that the welding torch can avoid obstacles while maintaining welding operation capability; The priority order of the obstacle avoidance adjustment strategies is as follows: a combination of deceleration strategy and attitude adjustment strategy is preferred; when attitude adjustment cannot meet the obstacle avoidance requirements, an offset strategy is activated; and when the offset strategy still cannot avoid the obstacle, a stopping strategy is activated. The dynamic adjustment of welding parameters in step S4 includes: during obstacle avoidance, dynamically adjusting the welding current and welding voltage based on the relative posture deviation between the welding torch and the weld to maintain a stable heat input and ensure the consistency of weld quality between the obstacle avoidance section and the non-obstacle avoidance section; after obstacle avoidance is completed, the digital twin model generates a transition path that smoothly returns from the current position to the preset welding path through a path interpolation algorithm, guiding the welding equipment to return to the planned position of the preset welding path along the transition path. The curvature of the transition path is continuous, and the rate of change of acceleration does not exceed a preset comfort threshold.

6. The intelligent obstacle avoidance method for gate welding based on digital twin technology according to claim 1, characterized in that, The obstacle avoidance data in step S5 includes the timestamp of the obstacle avoidance trigger time, the three-dimensional coordinates of the obstacle avoidance trigger position, the path trajectory point sequence during the obstacle avoidance execution process, the welding parameter adjustment record during the obstacle avoidance process, and the timestamp of the obstacle avoidance completion time; the path adjustment data includes the deviation data and deviation correction amount between the initial welding path and the actual execution path after obstacle avoidance; the welding quality inspection data includes weld forming size inspection data and weld internal defect inspection data; the updating of model parameters and obstacle avoidance strategy library specifically includes: performing cluster analysis on multiple obstacle avoidance data through machine learning algorithms to identify high-frequency obstacle avoidance areas and high-frequency obstacle avoidance types, optimizing the obstacle avoidance path templates and parameter matching rules in the obstacle avoidance strategy library, and incrementally updating the obstacle space occupancy information in the full-element digital twin model; The obstacle avoidance strategy library includes a process standard template library and an interference fault diagnosis template library. It simulates and extrapolates the motion trajectory of welding equipment and the spatial positional relationship between the welding torch sweeping space and obstacles under various working conditions. It collects spatial interaction data of each simulation node and simultaneously simulates the welding temperature field and stress field distribution under the corresponding working conditions. It analyzes the correlation between trajectory interference, component deformation and weld defects, generates standardized interference-free welding path samples and interference fault diagnosis reports, and stores them in the process standard template library and the interference fault diagnosis template library, respectively.

7. The intelligent obstacle avoidance method for gate welding based on digital twin technology according to claim 1, characterized in that, The intelligent obstacle avoidance method for gate welding also includes an obstacle avoidance priority arbitration step: when the welding quality maintenance requirement conflicts with the obstacle avoidance safety requirement, arbitration is carried out based on a preset priority rule. The priority rule is the safety priority principle, that is, ensuring the safety of the welding equipment and gate components is the primary goal, and maximizing the stability of welding quality under the premise that safety is guaranteed; when multiple obstacles exist at the same time and the obstacle avoidance strategies conflict with each other, priority is ranked based on the degree of danger of each obstacle, and the obstacle with the highest degree of danger is avoided first.

8. The intelligent obstacle avoidance method for gate welding based on digital twin technology according to claim 1, characterized in that, The welding equipment is a nine-axis linkage welding robot, which includes the robot body and two to seven additional motion axes. In step S2, when generating the collision-free welding path, the digital twin model is imported into the robot joint kinematics constraint model to simultaneously verify the travel limits and rotation limits of each joint and additional axis, and generate the optimal obstacle avoidance welding path that simultaneously satisfies the axis motion limit constraints and obstacle space avoidance constraints.

9. A gate welding intelligent obstacle avoidance system based on digital twin technology, characterized in that, include: The model building module is used to collect three-dimensional design data of gate components, establish a full-element digital twin model containing geometric structure information, material property information, assembly relationship information and welding process parameter information, and build a virtual welding equipment model and a virtual working environment model in the simulation platform based on the kinematic parameters of the welding equipment. The path planning and pre-simulation module is used to automatically extract weld feature information based on the full-element digital twin model and generate an initial welding path and initial welding torch posture. In the simulation platform, it drives the virtual welding equipment model to move along the initial welding path to perform collision avoidance simulation analysis, calculates the spatial distance between each component in real time and determines the collision risk, and generates a collision-free welding path and corresponding obstacle avoidance process scheme through iterative optimization. The multi-source sensor perception module includes a vision sensor, a distance sensor, and a force sensor installed on the welding equipment. It is used to collect perception data of the welding operation area in real time during the welding process, identify static obstacle information and dynamic obstacle information based on the perception data, and calculate the relative pose relationship between the welding torch and each obstacle in real time. The dynamic obstacle avoidance and replanning module is used to map real-time sensing data to the full-element digital twin model after coordinate transformation, so that the pose of the virtual welding equipment is synchronized with the actual pose of the physical welding equipment. When an obstacle is detected to enter the preset safe distance range or a collision is predicted to occur, an obstacle avoidance adjustment strategy is generated to dynamically adjust the welding torch movement trajectory and / or welding parameters, control the welding equipment to perform obstacle avoidance actions, and automatically return to the planned position of the preset welding path after obstacle avoidance is completed. The closed-loop feedback and update module is used to feed back obstacle avoidance data, path adjustment data and welding quality inspection data generated during the actual welding process to the full-element digital twin model, and update the model parameters and obstacle avoidance strategy library through machine learning algorithms.

10. The intelligent obstacle avoidance system for gate welding based on digital twin technology according to claim 9, characterized in that, In the multi-source sensor perception module: the vision sensors include a 3D structured light vision sensor and a laser contour scanning sensor, used to collect three-dimensional point cloud data of the welding operation area and two-dimensional cross-sectional data of the weld contour, respectively; the distance sensors include an ultrasonic sensor and a laser ranging sensor, with a measurement accuracy of not less than ±0.5mm; the force sensor is a six-dimensional force / torque sensor, set at the connection flange between the welding torch and the welding robot body, with a sampling frequency of not less than 1000Hz; in the dynamic obstacle avoidance and replanning module, the virtual and actual pose synchronization frequency is not less than 10Hz, and the pose synchronization delay does not exceed 50ms; the obstacle avoidance adjustment strategy includes one or more combinations of deceleration strategy, stopping strategy, offset strategy, and attitude adjustment strategy, with the priority order of obstacle avoidance adjustment strategy as follows: the combination of deceleration strategy and attitude adjustment strategy is preferred, followed by offset strategy, and finally stopping strategy; The system also includes an obstacle avoidance priority arbitration module, which arbitrates the conflict between welding quality maintenance requirements and obstacle avoidance safety requirements based on preset priority rules. The priority rules prioritize the safety of welding equipment and gate components. When multiple obstacles exist simultaneously and obstacle avoidance strategies conflict with each other, the system prioritizes the obstacles based on their degree of danger, avoiding the obstacle with the highest degree of danger first. The closed-loop feedback and update module uses a convolutional neural network to perform feature extraction and cluster analysis on multiple obstacle avoidance data, identifies high-frequency obstacle avoidance areas and high-frequency obstacle avoidance types, and optimizes the obstacle avoidance path templates and parameter matching rules in the obstacle avoidance strategy library. The dynamic obstacle avoidance and replanning module has a built-in interference warning unit, which is used to divide the interference warning level into three levels based on the predicted minimum distance between the welding torch and the obstacle: Level 1 is slight deviation, Level 2 is moderate interference risk and Level 3 is severe interference risk, and outputs deviation record, trajectory fine-tuning and emergency stop command respectively. The system also includes an XR visualization rendering unit, which is used to render the sweep space three-dimensional envelope of the moving parts of the welding equipment, the obstacle three-dimensional model, the obstacle avoidance path and the spatial safety gap value in real time, and push the rendered image to the on-site monitoring terminal in real time. The path planning and pre-simulation module has a built-in kinematic constraint verification unit, which is used to import the stroke limit parameters of each joint and additional axis of the welding equipment, and output the final interference-free obstacle avoidance path after verifying the axis motion limit constraints.

Images