Friction stir welding seam intelligent correction method based on visual identification

Abstract

The invention provides a friction stir welding seam intelligent correction method based on visual identification, and relates to the technical field of friction welding. Image information of a welding seam area of a to-be-welded workpiece is collected in real time, image feature points are recognized, and the spatial pose of the welding seam area is obtained through a spatial transformation model; based on the space position deviation of the actual space pose and the theoretical welding track, a dynamic deviation correction path is generated; and a multi-point iterative optimization algorithm of motion control parameters and a real-time feedback control method are further adopted, so that high-precision real-time adjustment and dynamic compensation of the welding seam track are realized. According to the method, the accuracy and the stability of friction welding seam deviation rectification are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of friction welding technology, and more specifically, to a method for intelligent correction of weld seams in friction stir welding based on visual recognition. Background Technology

[0002] Welding, as a key processing technology in modern manufacturing, is widely used in aerospace, automotive manufacturing, shipbuilding, rail transportation, machinery and equipment, electronic components, and many other fields. With the continuous improvement of industrial automation and intelligent manufacturing, higher demands are being placed on the precision, stability, and automation level of welding processes. In actual welding production, errors exist in the processing, clamping, and handling of the workpieces to be welded, and factors such as material thermal deformation and mechanical vibration during welding inevitably lead to weld position deviations. If these weld deviations are not detected and compensated for in a timely and accurate manner, they will not only reduce welding quality but may also lead to increased product defect rates, decreased production efficiency, and even subsequent assembly problems, seriously affecting product reliability and consistency.

[0003] To address weld misalignment, traditional welding equipment typically relies on manual observation and adjustment, or automated control with pre-set fixed paths, for welding positioning and compensation. This method is highly dependent on operator experience, making real-time performance and accuracy difficult to guarantee. Furthermore, manual operation cannot meet the demands of modern production—high-paced, highly consistent, and highly reliable processes—and is increasingly ill-suited to the realities of complex working conditions.

[0004] In recent years, with the rapid development of machine vision technology, sensor technology, automatic control technology, and intelligent optimization algorithms, intelligent welding correction methods based on visual sensing technology have been gradually proposed and applied. Real-time acquisition of weld area images, using image processing algorithms to identify weld features and calculate weld spatial pose, and then achieving real-time weld correction through path planning and motion control, has become an important trend in the development of welding automation technology. However, existing technical solutions still have many shortcomings in practical applications, especially for high-precision weld correction scenarios. Achieving more accurate, efficient, and robust real-time correction control still faces many technical challenges and urgently requires further exploration and improvement. Summary of the Invention

[0005] To overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a method for intelligent correction of weld seams in friction stir welding based on visual recognition.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A visual recognition-based intelligent correction method for friction stir weld seams includes: Based on the image information of the weld seam area of ​​the workpiece to be welded, image feature points are identified, and the spatial pose of the weld seam area is generated through a spatial transformation model. Weld offset is generated based on the difference between the spatial pose and the theoretical welding trajectory, and a dynamic correction path for welding trajectory correction is generated simultaneously. The motion control parameters of the welding equipment are generated based on the dynamic correction path, and the motion control parameters are iteratively corrected at multiple points based on the weld offset. Weld seam correction is completed by adjusting the welding position in real time based on the iteratively corrected motion control parameters.

[0007] Furthermore, the spatial transformation model includes a binocular 3D reconstruction module and a rigid body transformation module, wherein: The binocular 3D reconstruction module is used to acquire left and right views of the weld area using a binocular industrial camera, calculate the disparity value through a stereo matching algorithm, and further calculate the initial spatial 3D coordinates corresponding to each 2D image feature point. The rigid body transformation module is used to transform the initial three-dimensional spatial coordinates to the workpiece coordinate system to obtain the spatial coordinates in the workpiece coordinate system, which is used to determine the spatial pose of the weld area.

[0008] Furthermore, the generation of the dynamic correction path and the generation of the weld offset are synchronously achieved based on the difference between the spatial pose and the theoretical welding trajectory, including: The path compensation amount is determined based on the position difference vector between the theoretical and actual weld trajectory positions, and a dynamic adjustment factor is used to weight the path compensation amount to obtain the coordinates of the correction path points; the calculation formula for the coordinates of the dynamic correction path points is as follows: ,in, Let be the three-dimensional spatial coordinates of the s-th correction path point in the dynamic correction path. Let be the three-dimensional spatial coordinates of the s-th position in the theoretical welding trajectory. Let be the position difference vector between the theoretical trajectory and the actual trajectory at the s-th position. , This is the path dynamic adjustment factor.

[0009] Furthermore, the multi-point iterative correction of the motion control parameters based on the weld offset includes: The dynamic correction path is divided into multiple sub-intervals. At the beginning of each sub-interval, the position deviation between the target trajectory point position on the dynamic correction path and the actual arrival position determined by the motion control parameters is calculated to obtain the position deviation dataset of the corresponding sub-interval. Based on the position deviation dataset, the motion control parameters are updated using a weighted least squares iterative algorithm until the position deviation meets the preset convergence condition, thus obtaining the corrected motion control parameters.

[0010] Furthermore, the method for determining the preset convergence condition includes: The position deviation vector magnitude of the position deviation dataset during the calculation iteration is compared with the position deviation vector magnitude of the position deviation dataset in the previous iteration; when the ratio is less than a preset convergence threshold, the iteration is determined to be converged.

[0011] Furthermore, the process of determining the updated weights includes: Based on the position deviation dataset, the motion control parameter update weights corresponding to each sub-interval are determined according to the proportional relationship between the position deviation vector magnitude in each sub-interval and the total position deviation vector magnitude of all sub-intervals. The update weights are positively correlated with the position deviation vector magnitude of the corresponding sub-interval.

[0012] Furthermore, the real-time adjustment of the welding position employs a motion adaptive control model, including: The positional deviation information between the current position of the weld and the corresponding position of the dynamic correction path is collected in real time. The positional deviation information is used as feedback input, and the motion control parameters of the welding equipment are adjusted in real time through a proportional-integral control algorithm. The adjustment amount of the motion control parameters is positively correlated with the amplitude of the positional deviation information.

[0013] Furthermore, the process of determining the position deviation information includes: The current position of the welding equipment is detected in real time, the deviation vector between the current position and the corresponding theoretical position on the dynamic correction path is calculated, and the projection of the deviation vector in the vertical direction of the weld is used as the position deviation information as feedback input.

[0014] Furthermore, the control output of the proportional-integral control algorithm is specifically determined by the following formula: , in, This represents the output value of the motion control parameters in the k-th control cycle. The feedback position deviation is the value in the k-th control cycle. , These are the proportional gain coefficient and the integral gain coefficient, respectively.

[0015] Furthermore, the method also includes: Establish the association between weld offset, final motion control parameters, weld area image and unique identifier of workpiece to be welded, and store the association in data storage unit; the association is called by production management system for production quality analysis and traceability management.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention accurately obtains the spatial pose of the weld area through visual recognition and binocular 3D reconstruction, effectively solving the technical problems of insufficient accuracy, poor real-time performance, and weak anti-interference ability of traditional weld position detection methods, and realizing high-precision, real-time accurate measurement of weld position.

[0017] This invention overcomes the shortcomings of traditional static or simple correction paths in adapting to workpiece deformation and processing errors by constructing a dynamic compensation model for weld position deviation and using a multi-point iterative optimization method for motion control parameters, thus effectively improving the accuracy and stability of the actual welding path.

[0018] This invention effectively eliminates the influence of dynamic disturbances in the actual welding process by introducing a proportional-integral closed-loop control algorithm based on real-time position feedback, significantly improving the response speed and control accuracy of weld seam correction, and ensuring the reliability and consistency of welding quality. Attached Figure Description

[0019] Figure 1 The flowchart illustrates a visual recognition-based intelligent correction method for friction stir welding weld seams, as provided by this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1 Please see Figure 1 As shown, this embodiment discloses a method for intelligent correction of weld seams in friction stir welding based on visual recognition, including: Based on the image information of the weld seam area of ​​the workpiece to be welded, image feature points are identified, and the spatial pose of the weld seam area is generated through a spatial transformation model. It should be noted that the image information is a two-dimensional image of the weld seam area of ​​the workpiece to be welded, which is acquired in real time by a vision sensor. The vision sensor includes, but is not limited to, a binocular industrial camera, a structured light camera, or a depth camera. In order to accurately obtain the depth information of the two-dimensional feature points, this embodiment uses a binocular industrial camera to capture images of the weld seam area simultaneously by the left and right cameras. The disparity map is calculated based on a binocular stereo matching algorithm (such as the semi-global matching SGM algorithm), and the initial three-dimensional coordinates of the two-dimensional image feature points are further calculated.

[0022] It should be understood that the method for identifying image feature points can employ image feature detection algorithms, specifically including but not limited to Harris corner detection, Scale Invariant Feature Transform (SIFT), and ORB algorithm; this embodiment employs the SIFT algorithm, and its detection process includes: A scale-space image is constructed, extreme key points are identified, and then these key points are finely adjusted in position and subjected to stability screening to determine the final set of two-dimensional coordinates of image feature points, specifically represented as follows: , in, A set of two-dimensional coordinates of image feature points. Let be the two-dimensional coordinates of the i-th feature point in the image coordinate system, and N be the total number of feature points.

[0023] In implementation, the spatial transformation model includes a binocular 3D reconstruction module and a rigid body transformation module, wherein: The binocular 3D reconstruction module is used to acquire left and right views of the weld area using a binocular industrial camera, calculate the disparity value through a stereo matching algorithm, and further calculate the initial spatial 3D coordinates corresponding to each 2D image feature point. The rigid body transformation module is used to transform the initial three-dimensional spatial coordinates to the workpiece coordinate system to obtain the spatial coordinates in the workpiece coordinate system, which is used to determine the spatial pose of the weld area.

[0024] Specifically, the binocular 3D reconstruction module acquires left and right views of the weld seam area of ​​the workpiece to be welded using left and right cameras, obtaining a set of feature points for the left and right images: ; in, , This represents the pixel coordinates of the i-th feature point in the two-dimensional image coordinate system of the left camera. , This represents the pixel coordinates of the i-th feature point in the two-dimensional image coordinate system of the right-side camera; It should be noted that the feature point matching method for the left and right views in this embodiment is as follows: First, the coordinates of two-dimensional feature points in the left and right images are extracted using the SIFT algorithm, and a 128-dimensional descriptor vector corresponding to each feature point is obtained; then, nearest neighbor matching is performed based on the feature point descriptors using Euclidean distance, thereby establishing a one-to-one matching relationship between corresponding feature points in the left and right images, and the disparity value of each feature point is calculated based on this matching result; in addition, the complete disparity map of the weld area can also be calculated first using the semi-global matching (SGM) algorithm, and then the corresponding disparity map position can be directly queried using the two-dimensional coordinates of each SIFT feature point, thereby quickly obtaining the disparity value of each feature point.

[0025] The disparity value of each feature point is calculated using a stereo matching algorithm (such as the semi-global matching SGM algorithm). The specific calculation formula is as follows: ; In the formula, Let be the disparity value of the i-th feature point, representing the difference in horizontal pixel coordinates of the same feature point in the left and right camera images; Based on the binocular vision imaging principle and the calibrated camera intrinsic and extrinsic parameters, the depth information of this feature point is calculated using the following formula: ; In the formula, f is the actual physical focal length of the binocular camera lens, representing the distance from the optical center of the camera lens to the imaging sensor; B is the baseline length of the binocular camera, representing the distance between the optical centers of the left and right cameras. Furthermore, the three-dimensional coordinates of this feature point in the initial spatial coordinate system are calculated as follows: , in, , The principal point coordinates of the camera represent the position of the intersection of the camera's optical axis and the imaging sensor in the image coordinate system. , These are the pixel focal lengths of the camera in the horizontal and vertical directions, respectively, which are the ratios of the lens focal length to the pixel size of the imaging sensor.

[0026] The rigid body transformation module transforms the initial spatial coordinates using rotation matrices and translation vectors. Transform to the workpiece coordinate system to obtain the spatial coordinates in the workpiece coordinate system. The specific formula is as follows: , In the formula, For rotation matrix, It is the displacement vector; It should be noted that the rotation matrix and displacement vector are determined through a coordinate system calibration experiment. The calibration experiment specifically includes: measuring the three-dimensional spatial coordinates of multiple calibration feature points in the initial spatial coordinate system and the workpiece coordinate system respectively; calculating the two sets of spatial coordinates using a coordinate system matching algorithm (including but not limited to least squares fitting or singular value decomposition method) to determine the optimal transformation relationship between the two coordinate systems, thereby obtaining the specific values ​​of the corresponding rotation matrix and displacement vector.

[0027] After processing by the above two modules, the final set of spatial coordinates in the workpiece coordinate system is obtained, specifically represented as follows: , Based on the set of spatial coordinates in the workpiece coordinate system, a fitting algorithm is used to determine the spatial pose of the weld area of ​​the workpiece to be welded. It should be noted that the fitting algorithm includes, but is not limited to, least squares fitting, singular value decomposition, etc. In this embodiment, spatial spline curve fitting is used to obtain a continuous three-dimensional curve representing the actual weld trajectory. Based on this continuous curve, a local coordinate system is determined for each point along the weld, with its origin located at each point on the trajectory. The local coordinate axes are respectively composed of the tangential vector, normal vector, and binormal vector of the weld trajectory, thereby clearly giving the spatial pose sequence of the weld region along the entire trajectory. This spatial pose sequence is used for subsequent point-by-point comparison with the theoretical weld trajectory.

[0028] Weld offset is generated based on the difference between the spatial pose and the theoretical welding trajectory, and a dynamic correction path for welding trajectory correction is generated simultaneously. It should be noted that the theoretical welding trajectory is an ideal welding path pre-planned and stored in the welding equipment control system according to welding process requirements and workpiece design drawings; the theoretical welding trajectory is specifically stored in the form of a series of discrete three-dimensional coordinate points, which can be represented as: , In the formula, Let be the three-dimensional coordinates of the s-th point in the theoretical welding trajectory, and M be the total number of theoretical trajectory points.

[0029] It should be noted that, in order to ensure a one-to-one correspondence between the theoretical welding trajectory points and the actual weld trajectory points, this embodiment uses a nearest neighbor search algorithm to determine the theoretical trajectory points. In the actual weld trajectory Corresponding points on This allows for the calculation of the spatial position deviation vector. The specific calculation formula is as follows:

[0030] In the formula:

[0031] in, This represents the three-dimensional spatial deviation vector between the theoretical trajectory point and its nearest neighbor point on the actual weld trajectory; Points on the actual weld trajectory and the theoretical trajectory The nearest spatial point; This represents the Euclidean distance between two points.

[0032] It should be noted that, due to the actual weld trajectory For continuous fitting curves, while theoretical welding trajectories... Given a discrete set of points, to clarify the correspondence between the points and the theoretical trajectory, this embodiment uses a nearest neighbor search algorithm to determine the correspondence between the points, that is, for each point on the theoretical trajectory... In the actual weld trajectory Find the point in the middle that is spatially closest to the actual trajectory, and use it as the corresponding point to calculate the weld offset. .

[0033] The dynamic correction path is a trajectory for motion control of actual welding equipment, obtained by dynamically compensating and correcting the weld offset based on the theoretical trajectory. It can be expressed as: ; Specifically, the generation of the dynamic correction path and the generation of the weld offset are synchronously achieved based on the difference between the spatial pose and the theoretical welding trajectory, including: The path compensation amount is determined based on the position difference vector between the theoretical and actual weld trajectory positions, and a dynamic adjustment factor is used to weight the path compensation amount to obtain the coordinates of the correction path points; the calculation formula for the coordinates of the dynamic correction path points is as follows: ,in, Let be the three-dimensional spatial coordinates of the s-th correction path point in the dynamic correction path. Let be the three-dimensional spatial coordinates of the s-th position in the theoretical welding trajectory. Let be the position difference vector between the theoretical trajectory and the actual trajectory at the s-th position. , This is the path dynamic adjustment factor.

[0034] It should be noted that the path dynamic adjustment factor is used to perform nonlinear weighted adjustment of the path compensation amount, wherein... Used to adjust the overall weight of path compensation. The nonlinear attenuation trend used to adjust the sensitivity of path compensation to positional deviation can be obtained by those skilled in the art through conventional parameter tuning based on the material of the workpiece to be welded, the response characteristics of the welding equipment, and the welding accuracy requirements. In this embodiment, the adjustment factor... The possible value range is 0.5 to 2.0. The value range is 0.1 to 1.0.

[0035] It should also be noted that, to avoid over-adjustment or jitter caused by minor deviations, the exponential term in the formula... The design is to moderately compress the compensation amount when the deviation is small, thereby reducing the impact of small disturbances; when the deviation is large, the compensation tends to be fully corrected, achieving rapid deviation elimination.

[0036] The motion control parameters of the welding equipment are generated based on the dynamic correction path, and the motion control parameters are iteratively corrected at multiple points based on the weld offset. It is understood that the initial iterative optimization process of the motion control parameters belongs to the offline simulation learning stage. Its function is to pre-optimize the set of target trajectory point positions for the welding equipment tracking path based on a given dynamic correction path, using an iterative learning control method to compensate for the dynamic tracking error of the equipment itself, thereby obtaining an optimal set of target trajectory point coordinates. This optimized set of target trajectory point coordinates is used to replace the original dynamic correction path and serve as the benchmark for the optimized path of the welding equipment tracking in the actual welding process. Meanwhile, in the actual welding process, a closed-loop control loop based on real-time position feedback is also set up to fine-tune and compensate the initial parameters in real time according to the detected position deviation information, thereby realizing dynamic correction of the actual weld position error; the two control loops cooperate with each other to ensure the accuracy and robustness of the weld correction process.

[0037] It should be noted that, in this embodiment, "motion control parameters" refers to the set of control command variables upon which the welding equipment executes the dynamic correction path. This set describes the target motion state of the welding equipment in three-dimensional space, and its contents include the position coordinates, motion velocity, and acceleration information of the target trajectory points.

[0038] In the offline optimization phase, the spatial coordinates of the target trajectory points are iteratively optimized to correct the trajectory tracking error of the welding equipment, thereby obtaining the optimal set of trajectory points for subsequent real-time control. The target trajectory points can be represented as: , in, Let Q be the three-dimensional spatial coordinates of the target trajectory point in the j-th control cycle, and let Q be the total number of trajectory points. It is understandable that when the welding equipment executes the aforementioned set of trajectory points, its corresponding velocity, acceleration, and controller gain parameters can be determined through conventional debugging methods of the equipment control system and are not optimization variables of this offline optimization algorithm. Therefore, the goal of offline optimization is to finely correct the spatial position parameters to improve the path tracking accuracy in the subsequent online control stage.

[0039] It should be noted that the target trajectory point set finally obtained through the offline iterative optimization algorithm will replace the initial dynamic correction path and serve as the reference for the actual welding path in the online real-time control process. The position deviation information fed back by the online real-time control process is also determined based on the corresponding position in the above-mentioned optimized target trajectory point set, thereby ensuring that the offline optimization results can be effectively utilized by the online real-time control to achieve accurate weld trajectory tracking and correction.

[0040] In implementation, the multi-point iterative correction of motion control parameters based on weld offset includes: The dynamic correction path is divided into multiple sub-intervals. At the beginning of each sub-interval, the position deviation between the target trajectory point position on the dynamic correction path and the actual arrival position determined by the motion control parameters is calculated to obtain the position deviation dataset of the corresponding sub-interval. Based on the position deviation dataset, the motion control parameters are updated using a weighted least squares iterative algorithm until the position deviation meets the preset convergence condition, thus obtaining the corrected motion control parameters.

[0041] During implementation, the dynamic correction path is divided into K sub-intervals, and the coordinates of the target trajectory point corresponding to the starting position of the dynamic correction path in each sub-interval are: The coordinates of the actual trajectory points reached by the welding equipment in each sub-section when executing the motion according to the existing motion control parameters are: Then the position deviation vector of the starting position of each sub-interval can be expressed as: ; Based on the position deviation vector of the starting position of each sub-interval, a position deviation dataset for the corresponding sub-interval is constructed, which can be represented as: , It should be noted that the reference for calculating the position deviation vector is the coordinates of the target trajectory point on the dynamic correction path. This is to ensure that the position deviation calculation is completely consistent with the actual correction target.

[0042] In this example, the motion control parameters are updated using a least-squares iterative algorithm. By optimizing the position deviation dataset for each sub-interval, the actual motion trajectory of the welding equipment gradually approximates the dynamic correction path. The objective function in this iterative process can be expressed as: , In the formula, U is the vector of motion control parameters to be optimized. The updated weight for the k-th sub-interval. This is the difference vector between the actual position and the theoretical position corresponding to the k-th sub-interval; By analyzing the objective function The solution is obtained through iterative steps, with the motion control parameters U updated in each iteration, until convergence.

[0043] The method for determining the preset convergence condition includes: The position deviation vector magnitude of the position deviation dataset during the calculation iteration is compared with the position deviation vector magnitude of the position deviation dataset in the previous iteration; when the ratio is less than a preset convergence threshold, the iteration is determined to be converged.

[0044] During implementation, the preset convergence condition is used to determine whether the iteration has reached a stable state, that is, to determine whether the motion control parameters have been optimized to meet the accuracy requirements; when the convergence condition is met, the iteration process terminates and the final motion control parameters are output.

[0045] During the iteration process, the position deviation vector of the k-th iteration in each sub-interval is: The magnitude of the corresponding position deviation vector can be expressed as: ; To determine the convergence of the iteration globally, it is necessary to sum or take a weighted average of the magnitudes of the position deviation vectors across all sub-intervals to obtain the total iteration deviation. ; Calculate the ratio of the magnitude of the position deviation vector in two consecutive iterations of the position deviation dataset. This ratio can be expressed as: , Where r is the ratio of the magnitude of the position deviation vector in two adjacent iterations. The sum of squared positional deviations calculated in the h-th iteration. For the first The sum of squared positional deviations obtained from each iteration; The ratio of the magnitude of the position deviation vector in two consecutive iterations is compared with a preset convergence threshold. In comparison, if it is less than or equal to the preset convergence threshold If the iteration converges, the iteration stops, and the calculated motion control parameters are the final parameters.

[0046] It should be noted that the preset convergence threshold is determined comprehensively based on the workpiece tolerance, the allowable range of weld deviation and the dynamic response time of the equipment, and the value range is 0.001 to 0.01. Those skilled in the art can obtain the optimal value through conventional calibration or simulation, which will not be elaborated on here.

[0047] Specifically, the process of determining the updated weights includes: Based on the position deviation dataset, the motion control parameter update weights corresponding to each sub-interval are determined according to the proportional relationship between the position deviation vector magnitude in each sub-interval and the total position deviation vector magnitude of all sub-intervals. The update weights are positively correlated with the position deviation vector magnitude of the corresponding sub-interval.

[0048] Understandably, since the dynamic correction path has been divided into multiple sub-intervals, the magnitude of the position deviation vector between the actual position and the theoretical position in each sub-interval is not the same. Therefore, it is necessary to determine the corresponding update weight of each sub-interval based on the magnitude of the position deviation vector of different sub-intervals, so as to make the correction process of the iterative algorithm more efficient and accurate.

[0049] The motion control parameter update weights can be expressed as: ; It should be understood that the larger the magnitude of the position deviation vector in a sub-interval, the higher the update weight of its corresponding motion control parameters. Thus, in the iterative process of the weighted least squares algorithm, the sub-intervals with large position deviations are corrected first, enabling the motion control parameters to be optimized quickly toward convergence.

[0050] Weld seam correction is completed by adjusting the welding position in real time based on the iteratively corrected motion control parameters.

[0051] It is understandable that in the actual welding process, the optimized target trajectory point set U_j is used as the welding path reference for online real-time control, and the motion trajectory of the welding equipment is dynamically corrected in real time to ensure that the welding equipment runs accurately along the optimized target trajectory and realizes real-time and accurate correction of the weld position. In order to ensure the accuracy and real-time performance of the weld correction process, this embodiment adopts a motion adaptive control model to dynamically adjust the welding position in real time.

[0052] In implementation, the real-time adjustment of the welding position adopts a motion adaptive control model, including: The positional deviation information between the current position of the weld and the corresponding position of the dynamic correction path is collected in real time. The positional deviation information is used as feedback input, and the motion control parameters of the welding equipment are adjusted in real time through a proportional-integral control algorithm. The adjustment amount of the motion control parameters is positively correlated with the amplitude of the positional deviation information.

[0053] It should be understood that the motion adaptive control model is a closed-loop control model based on a real-time feedback mechanism. This model measures the deviation information between the current position of the weld and the theoretical position in real time, and uses this deviation information as input feedback to the control system to adjust the motion control parameters of the welding equipment in real time, so as to dynamically compensate for and eliminate welding position errors. The magnitude of the adjustment of the motion control parameters in each control cycle is positively correlated with the magnitude of the real-time feedback position deviation (i.e., position error). This positive correlation means that when the feedback position deviation is large, the control system will automatically increase the adjustment magnitude of the motion control parameters to quickly reduce the weld position error; when the feedback position deviation is small, the control system will automatically decrease the adjustment magnitude of the motion control parameters to achieve precise fine-tuning and avoid over-adjustment or oscillation.

[0054] The process of determining the position deviation information includes: The current position of the welding equipment is detected in real time, the deviation vector between the current position and the corresponding theoretical position on the dynamic correction path is calculated, and the projection of the deviation vector in the vertical direction of the weld is used as the position deviation information as feedback input.

[0055] In practical implementation, real-time position detection is achieved by using a real-time position measuring device configured on the welding equipment, such as a high-precision encoder built into the servo motor, a laser displacement sensor, or a vision measurement system, to obtain the spatial position coordinates of the welding equipment in real time, which can be expressed as: , in, The three-dimensional coordinates of the current actual position of the welding equipment are measured in real time. The specific spatial coordinates of the current actual position of the welding equipment obtained through real-time measurement; Based on the optimized welding trajectory calculated from the actual weld space pose, the welding equipment control system calls the corresponding theoretical position coordinates from the optimized target trajectory point set in real time according to the current actual welding position, which can be expressed as: , in, These are the theoretical trajectory coordinates of the welding equipment's real-time position on the dynamic correction path. These are the spatial coordinates of the theoretical trajectory on the dynamic correction path. Specifically, the control output of the proportional-integral control algorithm is determined by the following formula: , in, This represents the output value of the motion control parameters in the k-th control cycle. The feedback position deviation is the value in the k-th control cycle. , These are the proportional gain coefficient and the integral gain coefficient, respectively.

[0056] It should be noted that the aforementioned proportional term The proportional term is used to respond in real time to the rate of change of position deviation; the faster or greater the deviation changes, the greater the adjustment provided by the integral term. This is used to continuously eliminate persistent small deviations in the weld position, and by accumulating historical deviation information, gradually adjust the weld position to the theoretical trajectory position; the proportional gain coefficient and integral gain coefficient The values ​​are obtained by those skilled in the art based on the actual welding equipment response speed, welding accuracy requirements, and control system stability through conventional debugging methods, and the range is [insert range here]. , ; Output This represents the correction amount for the target position coordinates in the current cycle. The correction amount will be directly superimposed on the target trajectory point coordinates corresponding to the current control cycle, so as to realize real-time compensation and fine-tuning of the weld position and realize immediate response to dynamic disturbances without changing the overall structure of the offline optimization path.

[0057] It is understandable that the correction amount output by the proportional-integral control algorithm... As a one-dimensional scalar, its value represents the position correction in the normal direction of the weld trajectory. In practical applications, the welding equipment determines the tangential direction along the weld trajectory, and the normal direction of the plane containing the weld is determined as the correction direction. Therefore, the target position in the current control cycle... Updated to: , in, The normal unit vector of the current weld position is calculated by the geometric relationship between the trajectory tangent vector after curve fitting and the workpiece coordinate system, thereby achieving precise position correction.

[0058] For example, suppose the proportional gain coefficient in a certain control cycle is... Integral gain coefficient , No. Periodic motion control parameter output , No. Position deviation of periodic feedback ; If the weld position deviation detected in real time during the k-th cycle is a large deviation... Then the real-time output value of the periodic motion control parameter is calculated as follows: ; At this time, the motion control parameters are adjusted in real time with an amplitude of [value missing]. ; If the weld position deviation detected in real time during the k-th cycle is a small deviation... Then the real-time output value of the periodic motion control parameter is calculated as follows: ; At this time, the motion control parameters are adjusted in real time with an amplitude of [value missing]. .

[0059] This embodiment achieves intelligent deviation correction of friction welds through a two-stage control mechanism combining visual recognition and dynamic compensation. Specifically, the visual recognition and pose determination stage provides three-dimensional spatial data of the weld area and generates a dynamic correction path; a multi-point iterative optimization algorithm is used to generate an initial set of globally optimized motion control parameters offline; and online real-time position correction and dynamic compensation control are performed based on real-time feedback of position deviation information.

[0060] Example 2 Based on Embodiment 1 above, this embodiment focuses on the effective storage, analysis, and traceability management of weld seam correction process data. The present invention provides a visual recognition-based intelligent correction method for friction stir weld seams, which further includes: Establish the association between weld offset, final motion control parameters, weld area image and unique identifier of workpiece to be welded, and store the association in data storage unit; the association is called by production management system for production quality analysis and traceability management.

[0061] It should be noted that the association relationship described in this embodiment refers to a clear data association and mapping relationship between important data generated during the actual welding process, including weld offset, motion control parameters finally executed by the welding equipment, and image data collected in the weld area, and the unique identifier of the specific workpiece to be welded (such as, but not limited to, QR code, serial number, production number), so as to facilitate subsequent analysis and management traceability of the production process.

[0062] Understandably, the purpose of establishing the above-mentioned relationships is to achieve effective preservation, accurate analysis, precise traceability and management of quality data during the welding production process; after the welding process is completed, the welding process data and weld quality information of the corresponding workpiece can be directly and quickly retrieved through the workpiece's unique identifier, providing reliable data support for product quality analysis, process optimization and quality problem traceability.

[0063] It should be understood that, in the specific implementation process, the weld offset is specifically the spatial position difference vector between the theoretical welding trajectory position and the actual weld trajectory position, the final motion control parameters are specifically the set of welding equipment motion control parameters determined after multi-point iterative correction and reaching the convergence condition, and the weld area image is specifically the two-dimensional image data of the weld area collected in real time by an industrial vision sensor.

[0064] In the specific implementation process, the unique identifier of the workpiece to be welded can be identified and recorded using a unique identification method such as a QR code, barcode, or production serial number, which are widely used in industry. In this embodiment, the QR code identification method is preferred. The information stored in the QR code includes, but is not limited to, the workpiece production date, batch number, model specifications, production line number, etc. The specific data encoding format can be determined according to the enterprise's MES (Manufacturing Execution System) database specifications.

[0065] The data structure of the association relationship is disclosed as an example below: , For example, the image data is named with the workpiece's unique identifier and the acquisition time, such as QRCODE12345678_20240924_100230.jpg, which represents a weld image acquired at 10:02:30 on September 24, 2024, with the workpiece's QR code being QRCODE12345678. The parameter data is stored in a database table, as shown in Table 1. Table 1: Parameter Data Table

[0066] During implementation, the production management system calls the associated data in the above data storage unit in real time through network interface or field industrial bus interface for real-time monitoring, statistical analysis and quality anomaly traceability management of welding production quality; for example, when weld quality problems occur, managers can directly call the corresponding welding data (weld offset, motion control parameters and weld image) through the workpiece QR code to quickly locate the problem and analyze the cause of production.

[0067] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

[0068] In conclusion, 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, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent correction of weld seams in friction stir welding based on visual recognition, characterized in that, include: Based on the image information of the weld seam area of ​​the workpiece to be welded, image feature points are identified, and the spatial pose of the weld seam area is generated through a spatial transformation model. Weld offset is generated based on the difference between the spatial pose and the theoretical welding trajectory, and a dynamic correction path for welding trajectory correction is generated simultaneously. The motion control parameters of the welding equipment are generated based on the dynamic correction path, and the motion control parameters are iteratively corrected at multiple points based on the weld offset. Weld seam correction is completed by adjusting the welding position in real time based on the iteratively corrected motion control parameters.

2. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 1, characterized in that, The spatial transformation model includes a binocular 3D reconstruction module and a rigid body transformation module, wherein: The binocular 3D reconstruction module is used to acquire left and right views of the weld area using a binocular industrial camera, calculate the disparity value through a stereo matching algorithm, and further calculate the initial spatial 3D coordinates corresponding to each 2D image feature point. The rigid body transformation module is used to transform the initial three-dimensional spatial coordinates to the workpiece coordinate system to obtain the spatial coordinates in the workpiece coordinate system, which is used to determine the spatial pose of the weld area.

3. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 1, characterized in that, The generation of the dynamic correction path and the generation of the weld offset are synchronously achieved based on the difference between the spatial pose and the theoretical welding trajectory, including: The path compensation amount is determined based on the position difference vector between the theoretical and actual weld trajectory positions, and a dynamic adjustment factor is used to weight the path compensation amount to obtain the coordinates of the correction path points; the calculation formula for the coordinates of the dynamic correction path points is as follows: in, Let be the three-dimensional spatial coordinates of the s-th correction path point in the dynamic correction path. Let be the three-dimensional spatial coordinates of the s-th position in the theoretical welding trajectory. Let be the position difference vector between the theoretical trajectory and the actual trajectory at the s-th position. , This is the path dynamic adjustment factor.

4. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 1, characterized in that, The multi-point iterative correction of motion control parameters based on weld offset includes: The dynamic correction path is divided into multiple sub-intervals. At the beginning of each sub-interval, the position deviation between the target trajectory point position on the dynamic correction path and the actual arrival position determined by the motion control parameters is calculated to obtain the position deviation dataset of the corresponding sub-interval. Based on the position deviation dataset, the motion control parameters are updated using a weighted least squares iterative algorithm until the position deviation meets the preset convergence condition, thus obtaining the corrected motion control parameters.

5. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 4, characterized in that, The method for determining the preset convergence condition includes: The position deviation vector magnitude of the position deviation dataset during the calculation iteration is compared with the position deviation vector magnitude of the position deviation dataset in the previous iteration; when the ratio is less than a preset convergence threshold, the iteration is determined to be converged.

6. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 4, characterized in that, The process of determining the updated weights includes: Based on the position deviation dataset, the motion control parameter update weights corresponding to each sub-interval are determined according to the proportional relationship between the position deviation vector magnitude in each sub-interval and the total position deviation vector magnitude of all sub-intervals. The update weights are positively correlated with the position deviation vector magnitude of the corresponding sub-interval.

7. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 1, characterized in that, The real-time adjustment of the welding position adopts a motion adaptive control model, including: The positional deviation information between the current position of the weld and the corresponding position of the dynamic correction path is collected in real time. The positional deviation information is used as feedback input, and the motion control parameters of the welding equipment are adjusted in real time through a proportional-integral control algorithm. The adjustment amount of the motion control parameters is positively correlated with the amplitude of the positional deviation information.

8. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 7, characterized in that, The process of determining the position deviation information includes: The current position of the welding equipment is detected in real time, the deviation vector between the current position and the corresponding theoretical position on the dynamic correction path is calculated, and the projection of the deviation vector in the vertical direction of the weld is used as the position deviation information as feedback input.

9. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 7, characterized in that, The control output of the proportional-integral control algorithm is specifically determined by the following formula: ， in, This represents the output value of the motion control parameters in the k-th control cycle. The feedback position deviation is the value in the k-th control cycle. , These are the proportional gain coefficient and the integral gain coefficient, respectively.

10. The intelligent correction method for friction stir welding seams based on visual recognition according to claim 1, characterized in that, The method further includes: Establish the association between weld offset, final motion control parameters, weld area image and unique identifier of workpiece to be welded, and store the association in data storage unit; the association is called by production management system for production quality analysis and traceability management.

Images