Industrial ct-based light beam automatic centering method and system

Abstract

The application relates to the technical field of industrial CT, and discloses a light beam automatic centering method and system based on industrial CT, which aims to solve the technical defects of low light beam and weld centering precision and poor reliability caused by thermal deformation in the welding process. The method comprises the following steps: before welding, obtaining three-dimensional CT body data of a to-be-welded workpiece to generate a nominal weld trajectory line; a thermal coupling finite element analysis model is established to simulate the welding process to generate a predicted thermal distortion field atlas; the predicted thermal distortion is superimposed on the nominal trajectory line to generate a dynamic welding trajectory; during the welding process, a laser profile measurement sensor is driven to collect three-dimensional profile data of a weld area in real time and extract an actual weld center position; and a Kalman filter state observer is established. By adopting the above technical scheme, the application can realize continuous and high-precision tracking of the weld center position, overcome the influence of thermal distortion, and significantly improve the quality and yield of welding of complex structural parts.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of nondestructive testing, and in particular, relates to a light beam automatic centering method and system based on industrial CT. BACKGROUND

[0002] The technical field of nondestructive testing includes industrial computer tomography, ultrasonic testing, radiographic testing, magnetic particle testing and other branches. The core content of this technical field is to check and test the internal and surface structures of materials, parts and equipment without damaging or affecting the use performance of the detected objects under the premise of using physical or chemical methods. Industrial computer tomography is a core means of obtaining three-dimensional structure information of an object by using X-rays to penetrate the object and being received by a detector, and then performing image reconstruction by a computer. With the improvement of precision manufacturing and quality control requirements, industrial CT technology has been widely used in the fields of aerospace, automobile industry and material science. The development of this technology not only promotes the precision of product quality testing, but also brings profound changes in reverse engineering and failure analysis.

[0003] Among them, the light beam automatic centering method refers to the accurate alignment between the X-ray source, the rotation center of the measured workpiece and the detector in the industrial CT scanning process through an automatic algorithm and a control system. The theme proposes an automatic solution for the time-consuming and unstable precision centering link in the pre-scanning preparation work. The method calculates the offset between the workpiece rotation center and the ideal scanning center by analyzing the features of the initial projection data or preview images, and drives the multi-axis motion platform to perform compensation correction. In this method, the system acquires a small amount of projection data, identifies the workpiece contour or specific markers using image processing algorithms, and calculates the accurate center position. This method aims to replace the traditional manual visual centering or calibration process relying on standard samples to improve the detection efficiency and image reconstruction quality.

[0004] The prior art mostly relies on manual adjustment by operators or semi-automatic calibration based on simple geometric models in the light beam centering process, lacks self-adaptive ability to complex workpiece morphology, and is difficult to guarantee high-precision centering consistency. For example, when facing asymmetric or complex internal structure workpieces, the center positioning deviation is caused by feature extraction errors based on edge detection algorithms. The centering process is mostly static single calibration, which cannot monitor and compensate the center deviation caused by equipment thermal drift or mechanical vibration in the scanning process in real time, resulting in motion artifacts and blurring in the reconstructed image. The centering algorithm has low utilization efficiency of projection data, usually only analyzes a single frame or a few frames of images, and fails to fully integrate all projection information to obtain a globally optimal solution, resulting in insufficient centering accuracy. The existing method does not establish a direct feedback mechanism for centering accuracy and image quality, so that the influence of centering error on the final detection result cannot be quantitatively evaluated. In the scene of precision welding, additive manufacturing and other high-resolution detection requirements, these problems are particularly prominent, directly affecting the defect detection rate and the accuracy of size measurement. SUMMARY

[0005] The purpose of the present application is to overcome the technical defects of low light beam and weld centering accuracy and poor reliability caused by thermal deformation in the welding process in the prior art, and to provide a light beam automatic centering method and system based on industrial CT.

[0006] In order to achieve the above purpose, the technical scheme adopted by the present application is as follows: a light beam automatic centering method based on industrial CT, comprising the following steps:

[0007] S1: obtaining three-dimensional industrial computed tomography (CT) volume data of a workpiece to be welded before welding, performing image segmentation processing on the three-dimensional industrial CT volume data to extract a three-dimensional point cloud model of a weld area, and calculating and generating a nominal weld trajectory line representing the weld geometry based on the three-dimensional point cloud model;

[0008] S2: establishing a thermal-mechanical coupled finite element analysis model matched with the material properties of the workpiece to be welded and the preset welding process parameters, inputting the nominal weld trajectory line as a heat source moving path into the thermal-mechanical coupled finite element analysis model, performing numerical simulation of the welding process, thereby calculating the four-dimensional time-varying displacement field of each spatial node on the workpiece caused by heat input in the welding process, and generating a predicted thermal distortion field map;

[0009] S3: superimposing the displacement vector corresponding to the point on the nominal weld trajectory line in the predicted thermal distortion field map on the spatial coordinates of the corresponding point on the nominal weld trajectory line to generate a dynamic welding trajectory compensated by thermal distortion;

[0010] S4: during the welding process, driving a laser profile measurement sensor with a fixed spatial positional relationship with the welding light beam to continuously collect real-time three-dimensional profile data of the weld area in front of the action point of the welding light beam at a preset sampling frequency, and extracting the current actual weld center position from the real-time three-dimensional profile data;

[0011] S5: establishing a Kalman filter state observer, taking the weld center position indicated by the dynamic welding trajectory at the current time as the predicted value of the system state, and taking the current actual weld center position extracted by the laser profile measurement sensor as the observed value of the system state, performing optimal state estimation through the Kalman filter state observer, calculating the deviation vector between the predicted position and the actual position, and generating a real-time position correction control signal for the welding light beam actuator according to the deviation vector.

[0012] As a further scheme of the present application, the step S1 of generating the nominal weld trajectory line specifically comprises:

[0013] S111: using a threshold-based region growing algorithm to process the three-dimensional industrial CT volume data, setting the lower threshold of the voxel gray value as the gray value corresponding to the material density of the workpiece, and the upper threshold as the gray value corresponding to the air density, identifying and segmenting the voxels of the weld gap and the bevel region to form independent weld geometry data;

[0014] S112: applying a three-dimensional surface extraction algorithm such as the Marching Cubes algorithm to the weld geometry data to generate a triangular mesh model representing the inner and outer surfaces of the weld, i.e. the three-dimensional point cloud model;

[0015] S113: performing slice processing on the three-dimensional point cloud model along the predetermined welding direction to obtain a series of two-dimensional weld cross-sectional profiles, and applying a geometric center calculation or bottom V-shaped tip identification algorithm to each two-dimensional weld cross-sectional profile to determine the center point coordinates of the section;

[0016] S114: connecting the center point coordinates of the series of two-dimensional weld cross-sectional profiles in three-dimensional space and fitting them using a B-spline curve or cubic spline interpolation algorithm to generate a smooth and continuous three-dimensional space curve, which is the nominal weld trajectory line.

[0017] As a further scheme of the present application, the step S2 of generating the predicted thermal distortion field map specifically comprises:

[0018] S211: Construct a geometric model of the thermal coupling finite element analysis model, which accurately reproduces the geometry of the workpiece and weld to be welded; according to the material specification of the workpiece, set the material attribute parameters of the model, including the thermal conductivity, specific heat capacity, density, elastic modulus, Poisson's ratio and thermal expansion coefficient changing with temperature;

[0019] S212: The heat source effect of the welding light beam is equivalent to a three-dimensional Gaussian distribution or a double-ellipsoid volume heat source model, and the mathematical expression is:

[0020]

[0021] Where P is the light beam power, η is the thermal efficiency, a, b, and c are the geometric size parameters of the heat source model, is the energy distribution coefficient of the front and rear ellipsoids, and the parameters are set according to the preset welding process;

[0022] S213: Discretize the nominal weld bead trajectory into a series of path points, and in the simulation, drive the volume heat source model to move along the path points at a preset welding speed; in each time step, solve the non-steady-state heat conduction equation;

[0023]

[0024] Calculate the three-dimensional temperature field distribution of the entire workpiece model;

[0025] S214: Input the three-dimensional temperature field calculated in each time step as a load into the structural mechanics analysis solver to solve the elastic-plastic constitutive relationship equation based on thermal strain increment , wherein is the stress increment, is the elastic-plastic matrix, is the total strain increment, is the thermal strain increment, thereby calculating the displacement vector of all nodes on the workpiece caused by the temperature field, and collecting the node displacement vectors of all time steps to form the predicted thermal distortion field map.

[0026] As a further scheme of the present application, the step S3 of generating a dynamic welding trajectory specifically comprises:

[0027] S311: Discretize the nominal weld bead trajectory into a series of path points, and in the simulation, drive the volume heat source model to move along the path points at a preset welding speed; in each time step, solve the non-steady-state heat conduction equation; Expressed as a function with arc length as a parameter, ;

[0028] S312: From the predicted thermal distortion field map, query and obtain the predicted three-dimensional displacement vector corresponding to the position when the welding is carried out to the arc length l , wherein t(l) is the time when the welding beam travels to the arc length l;

[0029] S313: vector addition of the displacement vector and the corresponding point coordinates on the nominal weld trajectory, to obtain the dynamic weld trajectory , whose calculation formula is ;

[0030] S314: discretization of the dynamic weld trajectory into a series of six-degree-of-freedom path points containing position and attitude information, with time stamps corresponding to the welding speed, forming an instruction sequence for direct execution by the welding robot controller.

[0031] As a further scheme of the present application, the step S4 of extracting the current actual weld center position specifically comprises:

[0032] S411: the laser profile measurement sensor consists of a line laser emitter and a high-speed industrial camera, the line laser emitter emits a fan-shaped laser beam with a center wavelength of 658 nanometers and a power of 50 milliwatts, which is irradiated on the weld area to form a laser line reflecting the weld cross-sectional profile; the high-speed industrial camera is equipped with a narrow-band filter with a center wavelength of 658 nanometers and a bandwidth of 10 nanometers, and images of the laser line are collected at a frame rate of 200 Hz;

[0033] S412: pre-processing of each collected image, including image denoising and binarization, and then extracting the center pixel coordinates of the laser line using the barycenter method or Steger algorithm to generate a two-dimensional point set , wherein and are pixel coordinates;

[0034] S413: according to the pre-calibrated camera intrinsic matrix and laser plane equation, the two-dimensional point set is converted from the pixel coordinate system to the three-dimensional space coordinate system under the weld robot base coordinate system, to obtain the real-time three-dimensional point cloud of the weld cross section ;

[0035] S414: feature recognition of the real-time three-dimensional point cloud, to determine the actual weld center position at the current time by searching for the lowest point in the point cloud or calculating the intersection point of the two slope lines of the V-shaped groove .

[0036] As a further scheme of the present application, the step S5 of calculating the deviation vector by the Kalman filter state observer specifically comprises:

[0037] S511: define the state vector of the system , wherein and respectively are the position deviations of the welding beam center in the transverse and longitudinal directions, and is the rate of change of the deviation;

[0038] S512: a prediction model of the system state is established, and the state transition equation is wherein is the prior state estimation at time k, is the posterior state estimation at time k-1, F is a state transition matrix, which is in the form of [[1, 0, Δt, 0], [0, 1, 0, Δt], [0, 0, 1, 0], [0, 0, 0, 1]], and Δt is a sampling time interval, is the process noise, and the covariance matrix Q represents the uncertainty of the prediction of the thermal distortion model;

[0039] S513: an observation model of the system is established, and the observation equation is wherein is the observation value at time k, i.e. the deviation between the actual welding center position measured by the laser profile measurement sensor and the welding center position indicated by the dynamic welding trajectory at the time; H is an observation matrix, which is in the form of [[1, 0, 0, 0], [0, 1, 0, 0]], is the observation noise, and the covariance matrix R is calibrated according to the measurement accuracy of the laser profile measurement sensor;

[0040] S514: the update step of Kalman filtering is performed, first, the Kalman gain is calculated wherein is the prior error covariance matrix; then the observation value is used to update the state estimation, and the posterior state estimation is obtained; finally, the error covariance matrix is updated;

[0041] S515: the deviation component is extracted from the posterior state of the optimal estimation as the deviation vector, which is transmitted to the underlying controller of the welding beam execution mechanism to generate a position correction signal superimposed on the main motion instruction.

[0042] A light beam automatic centering system based on industrial CT, for performing any of the foregoing methods, the system comprising:

[0043] a CT data processing module configured to obtain three-dimensional industrial CT volume data of a workpiece to be welded, generate a three-dimensional point cloud model of a welding seam area through image segmentation and three-dimensional reconstruction, and calculate a nominal welding seam trajectory line therefrom;

[0044] The welding process simulation module is configured to construct a thermal-mechanical coupling finite element analysis model matched with the workpiece, to simulate and calculate a four-dimensional time-varying displacement field of the workpiece caused by heat input according to preset welding process parameters and taking the nominal weld seam trajectory as a heat source path, and to generate a predicted thermal distortion field map;

[0045] The distortion compensation trajectory generation module is configured to superimpose corresponding displacement vectors in the predicted thermal distortion field map and the nominal weld seam trajectory in space and time to generate a dynamic welding trajectory compensated for thermal distortion;

[0046] The real-time geometry sensing module includes a laser profile measurement sensor fixedly connected with the welding light beam and is configured to collect weld seam cross-section profile data in front of the light beam in real time during welding and extract a current actual weld seam center position therefrom;

[0047] The state estimation and control module internally integrates a Kalman filter state observer and is configured to receive a predicted position indicated by the dynamic welding trajectory and an actual position measured by the real-time geometry sensing module, to accurately calculate a deviation vector between the two positions by fusing information of the two positions through state estimation, and to output the deviation vector as a real-time position correction control signal for a welding light beam actuator.

[0048] Compared with the prior art, the application has the following advantages and positive effects:

[0049] The application calculates complex and nonlinear workpiece distortion caused by heat input in the welding process in advance by establishing a high-precision weld seam geometry model using industrial CT data before welding and combining thermal-mechanical coupling finite element analysis, thereby generating a dynamic welding trajectory with prospective distortion compensation. This converts the control problem from a pure lagging feedback correction to a predictive control based on a physical model. In the welding execution stage, the application does not rely on traditional visual sensors which are easily disturbed by welding arc light and smoke, but uses laser profile measurement technology with strong anti-interference capability to obtain real-time and accurate geometry information of the weld seam. Most importantly, the application establishes a Kalman filter state observer to optimally fuse a predictive value based on a physical model and an observed value based on real-time sensing, which fully utilizes global and prospective information provided by offline simulation and corrects model uncertainty and random disturbances in real time through online measurement. This closed-loop control framework combining prediction and correction enables the light beam centering system to overcome the influence of significant thermal distortion in the welding process and realize continuous and high-precision tracking of the weld seam center position, thereby significantly improving the quality and yield of welding of complex structural parts. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1A whole flow schematic diagram of a light beam automatic centering method provided by an embodiment of the present application is provided.

[0051] Figure 2 A functional module schematic diagram of a light beam automatic centering system provided by an embodiment of the present application is provided.

[0052] Figure 3 A data flow schematic diagram of a dynamic welding trajectory generation process in an embodiment of the present application is provided.

[0053] Figure 4 A real-time closed-loop control flow schematic diagram based on Kalman filtering in an embodiment of the present application is provided.

[0054] Figure 5 An application scenario schematic diagram of a method provided by an embodiment of the present application is provided. DETAILED DESCRIPTION

[0055] To make the objectives, technical solutions and advantages of the present application clearer and more apparent, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that the specific embodiments described herein are only used to explain the present application, but not to limit the protection scope of the present application. In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only used to facilitate the description of the present application and simplify the description, and therefore cannot be understood as indicating or implying that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. In addition, in the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0056] In a specific embodiment, referring to Figure 1 The present application provides a light beam automatic centering method based on industrial CT, comprising the following steps:

[0057] Step S1, obtaining three-dimensional industrial computed tomography (CT) data of a workpiece to be welded before welding, performing image segmentation processing on the three-dimensional industrial CT data to extract a three-dimensional point cloud model of a weld area, and calculating and generating a nominal weld trajectory line representing the geometric shape of the weld based on the three-dimensional point cloud model.

[0058] Step S2, a thermal-mechanical coupled finite element analysis model matching the material properties of the workpiece to be welded and the preset welding process parameters is established, the nominal weld seam trajectory is input as the heat source moving path into the thermal-mechanical coupled finite element analysis model, and the numerical simulation of the welding process is carried out, so as to calculate the four-dimensional time-varying displacement field of each spatial node on the workpiece caused by heat input in the welding process, and generate a predicted thermal distortion field atlas.

[0059] Step S3, the displacement vector corresponding to the point on the nominal weld seam trajectory in the time sequence in the predicted thermal distortion field atlas is superimposed on the spatial coordinates of the corresponding point on the nominal weld seam trajectory, and a dynamic welding trajectory compensated by thermal distortion is generated. The data flow of this process is shown in FIG. 2. Figure 3

[0060] Step S4, during the welding process, a laser profile measurement sensor with a fixed spatial position relationship with the welding beam is driven to continuously collect real-time three-dimensional profile data of the weld seam area in front of the welding beam action point at a preset sampling frequency, and the current actual weld seam center position is extracted from the real-time three-dimensional profile data.

[0061] Step S5, a Kalman filter state observer is established, the weld seam center position indicated by the dynamic welding trajectory at the current time is taken as the predicted value of the system state, and the current actual weld seam center position extracted by the laser profile measurement sensor is taken as the observed value of the system state. The optimal state estimation is carried out through the Kalman filter state observer, the deviation vector between the predicted position and the actual position is calculated, and the real-time position correction control signal for the welding beam actuator is generated according to the deviation vector. The real-time closed-loop control process is shown in FIG. 3. Figure 4

[0062] Further, each of the above steps is described in detail.

[0063] The process of obtaining the three-dimensional industrial CT volume data of the workpiece to be welded and generating the nominal weld seam trajectory in step S1 can be divided into the following sub-steps.

[0064] ​​Step S111, a threshold-based region growing algorithm is used to process the three-dimensional industrial CT volume data. In one specific application scenario, the workpiece to be welded is a V-groove joint formed by butt joint of two 12mm-thick GH4169 high-temperature alloy plates. The industrial CT scanning system uses a 450kV micro-focus X-ray source, and the detector is a flat panel detector. During the scanning process, the tube voltage is set to 420kV, the tube current is 1.2mA, 2400 projection views are collected by 360-degree rotation, and the voxel size of the reconstructed three-dimensional volume data is 0.08mm x 0.08mm x 0.08mm. The volume data is represented as a three-dimensional array, and each element is a 16-bit gray value. The lower limit threshold and the upper limit threshold of the voxel gray value are based on the fact that the workpiece base material and the weld gap (usually filled with air or protective gas) have significant differences in X-ray attenuation coefficients, which are reflected as differences in gray values. Specifically, the upper limit threshold is determined by selecting multiple pure air regions in the volume data, calculating the distribution of the voxel gray values, taking the mean value plus three times the standard deviation as the result, for example, the gray value is 850. The lower limit threshold is determined by selecting multiple sample regions in the dense base material region away from the weld, calculating the distribution of the voxel gray values, taking the mean value minus three times the standard deviation as the result, for example, the gray value is 51200. Therefore, the voxels with gray values in the range [850, 51200] are identified as belonging to the weld gap and the groove region. The execution process of the region growing algorithm is as follows: first, a seed point is manually selected or automatically selected by image matrix analysis on any two-dimensional slice image containing the weld, and the seed point must be located inside the weld gap. Then, the algorithm takes the seed point as the starting point, checks the 26 adjacent voxels in the three-dimensional space, and if the gray value of the adjacent voxel is in the preset interval [850, 51200], it is added to the target region and used as a new seed point. The process is iterated until there is no new voxel that meets the condition can be added. Finally, the set of voxels included in the target region constitutes an independent weld geometry data.

[0065] As an alternative, image segmentation can also employ a deep learning based semantic segmentation network, such as a three-dimensional U-Net. This approach requires a pre-constructed annotated dataset containing CT volume data of various typical workpieces and the weld seam regions manually segmented by professionals as labels. The three-dimensional U-Net model is trained on this dataset to learn the mapping from the original CT voxel grayscale to "weld seam" or "non-weld seam". In actual application, the CT volume data of the workpiece to be welded is input into the trained three-dimensional U-Net model, which directly outputs a probability atlas with the same size as the original volume data, where the value of each voxel in the atlas represents the probability of belonging to the weld seam region. By setting a probability threshold (e.g. 0.95), the set of voxels with a probability higher than the threshold is identified as the weld seam geometry data. This approach has higher robustness when dealing with complex weld seam shapes or the presence of artifacts.

[0066] Step S112, a three-dimensional volume surface extraction algorithm is applied to the weld seam geometry data to generate a three-dimensional point cloud model. Specifically, the Marching Cubes algorithm is used. This algorithm processes each cube unit composed of 8 adjacent voxels. For each cube unit, according to whether its 8 vertices are located inside the segmented weld seam geometry data, 2^8 = 256 combinations can be formed. Each combination corresponds to a pre-set triangular patch topology. By querying a pre-computed lookup table, one or more triangular patches are determined to be generated inside the cube unit to approximate the isosurface. After traversing all cube units, the vertex set of all generated triangular patches constitutes the three-dimensional point cloud model representing the inner and outer surfaces of the weld seam. The point cloud density of this model is determined by the original CT voxel resolution. In one embodiment, the generated triangular mesh model contains about 2.5 million vertices and 5 million triangular patches.

[0067] Step S113, the three-dimensional point cloud model is sliced along a predetermined welding direction. The predetermined welding direction is usually defined as the main direction of the weld seam length extension, which can be specified by the operator or automatically determined by performing principal component analysis (PCA) on the point cloud model, where the first principal component direction is the welding direction. Set a slice interval, for example 0.5 mm, and generate a series of parallel slice planes along this direction. Each slice plane intersects the three-dimensional point cloud model to form a point set of a two-dimensional weld seam cross-sectional profile. For each two-dimensional cross-sectional profile, a geometric center calculation algorithm is applied to determine the center point coordinates. Specifically, for a closed profile composed of N points The calculation formula of the geometric center of the closed profile is:

[0068]

[0069] As an alternative, especially for V or U groove, a bottom V tip identification algorithm can be adopted. This algorithm first performs piecewise linear fitting on the 2D cross-sectional profile point set, and identifies the two main linear segments that constitute the groove sidewall. Then the intersection of the two linear segments is calculated, and this intersection is taken as the center point of the cross-section. This method is more suitable for scenarios that require the welding energy to be precisely applied to the root of the weld, as it directly locates the physical root position. In the process of processing, irregular profiles may be encountered due to CT noise or inaccurate segmentation. At this time, the RANSAC (Random Sample Consensus) algorithm can be applied to the 2D profile point set for linear or curve fitting to remove outliers and enhance the stability of feature point positioning.

[0070] Step S114, connecting and fitting the center point coordinates of the series of 2D weld cross-sectional profiles in three-dimensional space. Through step S113, an ordered three-dimensional point sequence is obtained;

[0071] where each point is the center point of the corresponding cross-section. To generate a smooth, continuous and differentiable trajectory, a cubic spline interpolation algorithm is used to fit the point sequence. This algorithm ensures that the generated curve not only passes through all data points, but also is continuous in first and second derivatives, which is crucial for ensuring the smoothness of the welding robot motion. The final generated three-dimensional space curve is the nominal weld trajectory, represented in parametric form where l is the arc length along the curve.

[0072] The process of establishing a thermal-mechanical coupled finite element model and generating a predicted thermal-induced distortion field map in step S2 is shown in FIG. 2. As shown in the function of the welding process simulation module, it can be further decomposed into the following sub-steps. Figure 2

[0073] Step S211, constructing the geometric model of the thermal-mechanical coupled finite element analysis model and setting material properties. The geometric model directly uses the three-dimensional model of the workpiece base part segmented in step S1, which accurately reproduces the macroscopic geometric form of the workpiece to be welded and the weld groove. The geometric model is meshed using unstructured tetrahedral elements. In order to balance the calculation accuracy and efficiency, a non-uniform mesh strategy is adopted. In the estimated heat affected zone (HAZ) and fusion zone, i.e. a columnar region with a radius of 5 mm around the nominal weld trajectory, the element characteristic size is set to 0.3 mm; in the base part away from the welding area, the element size gradually transitions to 3 mm. The final generated finite element mesh contains about 2.1x10^6 tetrahedral elements and 4.5x10^5 nodes.

[0074] ​Based on the material specifications of workpiece material GH4169 and publicly available academic literature databases, the temperature-dependent material property parameters in the model were set. The accuracy of these parameters directly affects the reliability of the simulation results. In a specific embodiment, the temperature-related thermophysical and mechanical property parameter data used are shown in the table below:

[0075] Temperature (°C) Thermal conductivity (W / (m-K)) Specific heat capacity (J / (kg-K)) Density (kg / m3) Elastic modulus (GPa) Poisson's ratio coefficient of thermal expansion (10⁻ 6 / K) Yield strength (MPa) 20 11.4 435 8240 205 0.294 12.8 1100 200 13.2 485 8190 198 0.301 13.5 1050 400 15.1 540 8130 189 0.309 14.2 1020 600 17.3 590 8060 175 0.318 15.1 980 800 20.5 650 7980 140 0.325 16.0 450 1000 24.1 710 7900 80 0.330 16.8 200 1200 28.0 760 7820 30 0.335 17.5 50 1400 32.5 810 7750 5 0.340 18.2 10

[0076] In the finite element solver, material property values ​​at any temperature are calculated using linear interpolation between these data points.

[0077] Step S212 involves mathematically modeling the heat source effect of the welding beam. The Goldak double-ellipsoidal volumetric heat source model is adopted because it can asymmetrically describe the temperature gradient before and after the weld pool, better reflecting the physical reality of high-speed welding. Its mathematical expression is:

[0078] For the first half of the ellipsoid (x≥0):

[0079]

[0080] For the posterior ellipsoid (x<0):

[0081]

[0082] Where P is the beam power and η is the thermal efficiency. , b and c are the geometric dimensions of the heat source model. And f_r are the energy distribution coefficients of the front and rear ellipsoids and + =2. These parameters are set according to the preset welding process. For example, for a laser filler wire welding process, the parameters are set as follows:

[0083] The beam power P = 4000 watts.

[0084] Welding speed v = 15 mm / s.

[0085] The heat input efficiency η = 0.70. This value is determined through calorimetric experiments or selected based on empirical data from similar processes.

[0086] length of the first hemisphere =2.5 mm.

[0087] Length of the posterior hemispheric =5.0 mm.

[0088] The weld width half-axis b = 2.0 mm.

[0089] The half-axis of the melt depth is c = 3.0 mm.

[0090] Energy distribution coefficient =0.6, =1.4.

[0091] The calibration of these geometric parameters is achieved by performing experimental welding on a test plate of the same material and thickness as the actual workpiece, then creating a metallographic section of the weld, measuring the actual size of the molten pool, and then adjusting the model parameters through reverse optimization to ensure that the error between the simulated molten pool morphology and the actual metallographic morphology (such as the error in weld width and weld depth) is less than 5%.

[0092] Step S213: Perform transient thermal analysis. First, the nominal weld trajectory line generated in step S114 is... The model is discretized into a series of path points. The distance Δl is determined by the welding speed v and the solution time step Δt, i.e., Δl = v * Δt. To ensure numerical convergence and accuracy, the time step Δt is set to 0.02 seconds, resulting in a path point spacing of 0.3 mm. During the simulation, the double ellipsoidal volumetric heat source model is driven to move along the discrete path points at a welding speed of 15 mm / s. Within each time step, the unsteady-state heat conduction equation is solved.

[0093]

[0094] in, , , These are density, specific heat capacity, and thermal conductivity as they change with temperature. The heat source term is a double ellipsoid. Boundary conditions are set to allow for both convection and radiation between all outer surfaces of the workpiece and the environment. The ambient temperature is set to 25℃, the convective heat transfer coefficient h is set to 20 W / (m²·K), and the surface emissivity ε is set to 0.8. By solving for the entire welding process and the subsequent cooling process (e.g., cooling to 500 seconds after welding begins), the three-dimensional temperature field distribution T(x,y,z,t) of all nodes in the workpiece model at each time point is calculated.

[0095] Step S214: Perform sequential coupled structural mechanics analysis. The transient temperature field T(x,y,z,t) calculated in step S213 is used as a thermal load and applied to the structural model at each time step. Before the mechanical analysis, mechanical boundary conditions need to be set. Based on the actual clamping conditions, three fixed constraint points are set at the bottom of the workpiece to restrict its rigid body displacement and rotation. In each incremental step, the elastoplastic constitutive relation equations based on the thermal strain increment are solved. .in For stress increment, To provide an elastic-plastic stiffness matrix that takes into account the effects of temperature, This represents the total strain increment. Thermal strain increment. according to Calculation, where temperature-dependent thermal expansion coefficient, temperature increment of the current step. The constitutive model adopted is a combination of von Mises yield criterion and kinematic hardening model, to more accurately describe the mechanical behavior of the material during cyclic heating and cooling. By solving the entire time sequence, the three-dimensional displacement vector of all finite element nodes on the workpiece at each time caused by the change of the temperature field is finally calculated . The collection of displacement vectors of all nodes at all time steps constitutes the predicted thermal distortion field map. This map is a high-dimensional data set, whose data structure is an associative array or hash table, with the key being the node ID and timestamp (node_id, t) and the value being the corresponding three-dimensional displacement vector [Ux, Uy, Uz].

[0096] As a more computationally efficient alternative, the inherent strain method can be used. This method first calculates the residual plastic strain distribution in a representative two-dimensional weld cross-section, i.e. the inherent strain, through a fine thermal-mechanical coupled analysis of the cross-section. Then, this two-dimensional inherent strain distribution is mapped along the nominal weld seam trajectory of the entire three-dimensional workpiece to form an initial strain field. Finally, only one linear elastic static structural analysis of the entire workpiece model is required, with the initial strain field as the load applied, to quickly calculate the final welding deformation. This method greatly reduces the calculation time, but sacrifices the accurate description of the dynamic deformation process during welding, and is suitable for scenarios where the final deformation state prediction is required to be highly accurate, while the process dynamic accuracy is not required to be high.

[0097] The process of generating the dynamic welding trajectory compensated for thermal distortion in step S3 has a data flow as shown in FIG. 11, and can be specifically decomposed into the following sub-steps. Figure 3

[0098] In step S311, the nominal weld seam trajectory is expressed as a function with arc length l as the parameter. The smooth curve generated in step S114 is calculated and stored as a sequence of densely sampled three-dimensional coordinate points, each point P_i being accompanied by a cumulative arc length value l_i from the starting point . For example, a weld seam with a length of 300 mm, if discretized with a precision of 0.1 mm, will generate a lookup table containing 3001 four-tuple (x, y, z, l). The coordinates of any intermediate position with arc length l can be obtained by linear or higher-order interpolation of the adjacent two points in the lookup table, thereby realizing the calculation expression of the function = [x(l), y(l), z(l)].

[0099] ​Step S312, from the predicted thermal distortion field map, the predicted three-dimensional displacement vector corresponding to the position at the arc length l is obtained. This query process contains matching and interpolation in both time and space. First, time matching is performed. The time t(l) corresponding to the arc length l is obtained by;

[0100]

[0101] is calculated, where v is a preset constant welding speed. For example, when the welding speed v is 15 mm / s, the time t(90) corresponding to the arc length l = 90 mm is 6.0 s. Second, spatial interpolation is performed. Since the coordinates of the point usually do not exactly coincide with the nodes of the finite element model, the displacement of the point needs to be calculated by interpolation according to the displacement values of the nodes around it. Specifically, first, the tetrahedral element containing the point is located in the finite element grid. This locating process can be accelerated by constructing a spatial index structure such as an octree. After finding the tetrahedral element, the displacement vectors U1, U2, U3, U4 of its four vertices (denoted as N1, N2, N3, N4) at the time t(l) can be directly queried from the predicted thermal distortion field map. The displacement vector of the point can be calculated by barycentric coordinate interpolation. If the barycentric coordinates of the point are (w1, w2, w3, w4), then its displacement vector is . As an alternative spatial interpolation scheme, the inverse distance weighted method (IDW) can be used. All k nodes within a radius R around the point are selected, and the displacement vectors of the nodes are weighted and averaged according to the inverse of the distance from the point . This process is repeated for each discrete point on the nominal weld trajectory to obtain a sequence of displacement vectors corresponding to each point.

[0102] Step S313, the displacement vector is vector added to the corresponding point coordinates on the nominal weld trajectory to obtain the dynamic weld trajectory. This calculation is performed independently for each discrete point on the trajectory. The calculation formula is . Taking a specific point as an example, assume that at the arc length l = 90 mm, the nominal trajectory point coordinates (90) are [150.2, 55.8, 12.0] (unit: mm). Through the query and interpolation of step S312, the predicted thermal displacement vector of the point at t = 6.0 s is ​(90,6.0) is [0.0,-0.45,0.25]. The coordinates of the corresponding point on the compensated dynamic welding trajectory are then (90) is [150.2, 55.8, 12.0] + [0.0,-0.45,0.25] = [150.2, 55.35, 12.25]. This calculation reveals that at this instant, the actual position of the weld has shifted -0.45 mm in the y direction (lateral) and +0.25 mm in the z direction (height) compared to its initial position. Performing this vector addition operation for all points on the nominal trajectory generates the complete set of dynamic welding trajectory points.

[0103] Step S314, discretize the dynamic welding trajectory into a sequence of six-DOF path points containing position and orientation information. This sequence is the instruction set that can be directly executed by the welding robot controller. For each dynamic trajectory point its position information [x, y, z] has been determined. Its orientation information, i.e. the direction of the welding beam axis, needs to be calculated additionally. In a typical embodiment, the beam axis should always be perpendicular to the workpiece surface and pointing towards the weld center. Therefore, the Z-axis direction of the tool coordinate system (usually the beam direction) can be determined jointly by the tangent direction of the dynamic trajectory at this point and the normal direction of the weld cross-section. Specifically, the tangent vector of the dynamic trajectory at point can be calculated by central difference method:

[0104]

[0105] This tangent vector defines one degree of freedom of the robot tool orientation (e.g. rotation around Y-axis). The other orientation vector, i.e. the X-axis direction of the tool coordinate system (lateral), can be defined as perpendicular to and lying in the local horizontal plane. By orthogonalizing these vectors, a complete six-DOF orientation (usually represented as Euler angles or quaternions) can be uniquely determined. The final generated instruction sequence is a list, each element of which contains: path point number, three-dimensional coordinates [x, y, z], three-dimensional orientation [Rx, Ry, Rz], linear velocity to be adopted to reach this point, and corresponding time stamp. This instruction sequence file is formatted into a specific robot controller compatible text format for easy loading and execution.

[0106] The process of acquiring three-dimensional profile data in real time during the welding process and extracting the actual weld center position in step S4 can be further decomposed into the following sub-steps.

[0107] ​Step S411, configure and use the laser profile measurement sensor. The sensor is physically fixed to the welding head by a precisely machined rigid bracket, ensuring the relative position and pose between the sensor and the welding beam focal point is constant. In a specific embodiment, the sensor lens optical axis is parallel to the welding beam axis, and the measurement point is located 20 mm in front of the welding beam action point. This "front distance" provides the necessary reaction time for the system. The fan-shaped laser beam emitted by the laser emitter has a central wavelength of 658 nm and a power of 50 mW. The reason for choosing a wavelength of 658 nm is that it is in the visible red light range, which is convenient for debugging observation, and is significantly different from the emission spectrum of common welding plasma or reflected laser, which is conducive to signal filtering. The high-speed industrial camera is equipped with a narrow-band interference filter whose central wavelength is accurately matched with the laser wavelength (658 nm), and its bandwidth (FWHM) is 10 nm. This extremely narrow bandwidth can effectively suppress the strong arc light, spatter and thermal radiation and other wide-spectrum background noise generated during welding, allowing only the reflected laser light to pass through, thereby significantly improving the signal-to-noise ratio of the image. The camera continuously acquires laser line images at a frame rate of 200 Hz, which is sufficient to capture the rapid dynamic changes in the weld position caused by thermal deformation, clamping stress release, etc., meeting the bandwidth requirements of real-time closed-loop control.

[0108] Step S412, pre-process and extract the laser line center for each frame of image collected. Each frame of original image is first subjected to a pre-processing procedure. In the image denoising stage, a 5x5 Gaussian filter is used to perform convolution operation on the image to smooth the random noise caused by sensor dark current or readout noise, while preserving the edge features of the laser line. Subsequently, binaryzation processing is performed, but not simple global threshold binaryzation, but an adaptive threshold method is used to deal with the problem of uneven brightness of the laser line under different substrate surface reflectivity. More preferably, no binaryzation is performed, and the center is directly extracted on the grayscale image. Specifically, the Steger algorithm is used to extract the sub-pixel center coordinates of the laser line. This algorithm first calculates the Hessian matrix of each pixel point in the image and solves its eigenvalues and eigenvectors. The center point of the laser line is defined as the point with the maximum second-order directional derivative perpendicular to the line direction. By solving a quadratic equation, the sub-pixel accurate position of these points can be analytically calculated. Compared with the barycenter method which can only provide pixel-level accuracy, the Steger algorithm can achieve an accuracy of 0.1 pixel or even higher, which is crucial for the accuracy of subsequent three-dimensional reconstruction. After this step, each frame of image is converted into a two-dimensional sub-pixel coordinate point set where i is from 1 to N, N is usually hundreds of points, which collectively constitute the projection of the laser line on the image sensor target surface.

[0109] Step S413, convert the two-dimensional pixel coordinate point set to three-dimensional space coordinates. This conversion relies on a pre-completed sensor calibration process. The calibration process includes intrinsic calibration and extrinsic (hand-eye) calibration. Intrinsic calibration uses a planar calibration board such as a checkerboard to calculate the intrinsic matrix K of the camera (including focal length , and principal point , ) and distortion coefficients by Zhang Zhengyou calibration method. Extrinsic calibration determines the equation of the laser plane in the camera coordinate system . For each extracted two-dimensional sub-pixel point , , first use the inverse of the intrinsic matrix K and the distortion model to project it back to a three-dimensional ray in the camera coordinate system. The direction vector of the ray is

[0110]

[0111] Then, calculate the intersection of the ray and the calibrated laser plane, and the three-dimensional coordinates of the intersection point is the true position of the point in the camera coordinate system. Finally, use the pre-calibrated homogeneous transformation matrix from the camera coordinate system to the welding robot base coordinate system to transform all three-dimensional points in the camera coordinate system to the robot base coordinate system to obtain the final real-time three-dimensional point cloud of the weld cross section . The entire conversion process is completed within each sampling period (5 milliseconds).

[0112] Step S414, feature recognition is performed on the real-time three-dimensional point cloud to determine the actual weld center position. For a typical V-shaped groove, a robust line fitting algorithm based on RANSAC is used. First, the acquired real-time three-dimensional point cloud The two-dimensional point set is projected onto a plane perpendicular to the welding direction. Then, two RANSAC line fittings are performed on the two-dimensional point set. The first fitting aims to identify the point set that constitutes the bevel wall of one side of the groove and obtain its line equation L_1. After removing the inliers belonging to L_1 from the point set, the second RANSAC line fitting is performed on the remaining point set to obtain the line equation L_2 of the bevel wall of the other side. The intersection point of the two lines L_1 and L_2 is accurately determined as the actual welding seam center (root) position Z_k at the current time. The advantage of this method is that the RANSAC algorithm can effectively resist the interference of local contour outliers caused by surface oxidation skin, temporary positioning welding points or a small amount of spatter, thereby providing a very stable and reliable welding seam center positioning result. For U-shaped or rectangular grooves, the feature recognition algorithm is adjusted accordingly to fit a curve or search for a symmetric center axis. The final output Z_k is a three-dimensional coordinate vector in the robot base coordinate system.

[0113] The process of calculating the bias vector in step S5 by the Kalman filter state observer is shown in the flowchart of Fig. 6. Specifically, it can be decomposed into the following sub-steps. Figure 4

[0114] Step S511, define the state vector of the system. The state vector is defined as . Wherein, and represent the position deviation of the welding beam center in the direction perpendicular to the welding direction (lateral direction) and along the beam axis direction (height or gap), respectively. These two directions constitute the key plane for centering control. and are the change rates of the deviation in these two directions, respectively. Including the velocity component in the state vector makes the filter not only estimate the current position deviation, but also predict the trend of the deviation at the next time, so as to realize smoother and more forward-looking control.

[0115] Step S512, establish the prediction model of the system state. This model describes how the system state naturally evolves over time. Its state transition equation is . is the prior state estimation at time k, that is, the optimal estimation ​Prediction of the current state. The state transition matrix F is constructed based on Newton's laws of kinematics, assuming that the rate of change of the deviation is approximately constant within a time interval as short as one sampling period Δt (e.g., 5 milliseconds). Therefore, the specific form of the F matrix is ​​[[1,0,Δt,0],[0,1,0,Δt],[0,0,1,0],[0,0,0,1]]. This matrix indicates that the current position deviation is equal to the position deviation at the previous moment plus the product of velocity and time, and the current velocity is the same as the velocity at the previous moment. This is the process noise term, which represents the uncertainty of the state model itself, mainly originating from the thermally induced distortion finite element model (…). The model incorporates prediction errors and unmodeled dynamic disturbances (such as small variations in workpiece clamping force). Process noise is assumed to be zero-mean Gaussian white noise, and its covariance matrix Q is a carefully tuned 4x4 matrix. The size of the diagonal elements of Q reflects the degree of confidence in the model's prediction bias and its rate of change. For example, if the simulation model's prediction accuracy in the lateral direction is lower than in the height direction, then the covariance matrix Q will contain elements that are less accurate in the lateral direction than in the height direction. and The corresponding diagonal element values ​​should be set larger to indicate greater uncertainty in that direction.

[0116] Step S513: Establish the system's observation model. This model connects the system state, which cannot be directly observed, with the directly measurable observables. Its observation equation is: The observed value z_k is the weld center position actually measured by the laser profile measurement sensor at time k. (From step S414) and the weld center position indicated by the dynamic welding trajectory at that moment. The difference vector between them, i.e. .because and They are all three-dimensional coordinates. It is also a three-dimensional vector, but we are only concerned with its components in the horizontal and vertical directions, therefore It is a two-dimensional vector [Δx, Δy]^T. The observation matrix H is used to extract the state vector from the four-dimensional vector. Extract the component directly corresponding to the observation. Under this definition, H has the form [[1,0,0,0],[0,1,0,0]], which directly extracts the position deviation component from the state vector. and Extract it. This is the observation noise term, representing the uncertainty in the measurement process. It mainly originates from the inherent measurement error of the laser profile measurement sensor, the accuracy limitations of the image processing algorithm, and the calibration error of the coordinate transformation. The observation noise is also assumed to be zero-mean Gaussian white noise, and its covariance matrix R is a 2x2 diagonal matrix. The diagonal elements of R... and are the quantified values of the variance of the sensor's measurements in the lateral and height directions. These values can be determined by offline experiments: align the sensor to a static, known-geometry V-groove standard block, and take repeated measurements for a long time (e.g. thousands of times), and then statistically determine the variances of the measurements in the lateral and height directions, i.e. to calibrate the R matrix.

[0117] Step S514, perform the recursive update step of Kalman filter. This step is the core of information fusion, and is executed in each sampling period.

[0118] First, predict the error covariance of the next state: This formula propagates the posterior error covariance from the last time step forward in time, using the state transition model, and adds the effect of process noise Q to get the prior error covariance .

[0119] Second, compute the Kalman gain: The Kalman gain is a 4x2 matrix, which is essentially a dynamically computed weight. It balances the uncertainty of the prior state estimate (reflected by ) and the uncertainty of the new observation (reflected by R). If the sensor is very accurate (R is small), will be large, making the filter trust the observation more. Conversely, if the physical model is very reliable ( is small), will be small, making the filter trust its own prediction more.

[0120] Then, update the state estimate using the observation: This is the correction step of the filter. is called the "innovation" or "measurement residual", and represents the difference between the actual measurement and the model-predicted value. This difference, multiplied by the Kalman gain , is used to correct the prior estimate , to get the optimal posterior state estimate .

[0121] Finally, update the error covariance matrix . This step computes the uncertainty of the system state estimate after incorporating the new observation information. The updated covariance is usually smaller than the prior covariance , quantitatively reflecting the precision improvement brought by information fusion. will be used as the .

[0122] Step S515, extract the deviation vector from the optimal estimated posterior state and generate a control signal. The optimal posterior state estimate is computed in step S514 Afterwards, the first two components are extracted directly, i.e. , as the current most reliable weld seam deviation vector. This deviation vector is transmitted to the motion controller at the base layer of the welding robot. Inside the controller, a separate, high-speed closed-loop position control algorithm is running, e.g. a proportional-integral-derivative (PID) controller. This PID controller takes the deviation vector as input, whose goal is to drive this deviation to zero. The controller outputs a correction velocity or correction displacement signal, which is superimposed in real-time with the main motion commands from P_dynamic that are being executed by the robot's main controller. For example, if the deviation vector is [-0.1, 0.05] (mm), the PID controller generates a tiny corrective motion pointing in the direction of [+0.1, -0.05], driving the welding beam back exactly to the actual weld seam center. This Kalman-filter-based prediction-correction control framework ensures a smooth, fast and accurate response of the entire system.

[0123] With reference to Figure 2 and Figure 5 , the present invention also provides an industrial CT-based beam auto-alignment system for performing the aforementioned method. This system is integrated into an automated welding workcell, which contains a six-axis industrial robot, a laser welding head, a wire feeding mechanism and corresponding workpiece fixtures. The system specifically comprises the following functional modules:

[0124] A CT data processing module, typically deployed on a high-performance workstation. This module is configured to receive raw 3D volume data files (e.g. in DICOM format) from an industrial CT device, and perform all the operations described in steps S111 to S114. Specifically, it internally integrates threshold-based region growing or 3D U-Net segmentation algorithms, marching cubes algorithm, multi-slice analysis and B-spline interpolation algorithms. The final output of this module is a nominal weld seam trajectory file, which characterizes the original geometry of the weld seam.

[0125] A welding process simulation module, also deployed on a high-performance computing server or cluster. The core of this module is a commercial or open-source finite element analysis software solver. This module is configured to receive the workpiece geometry model and the nominal weld seam trajectory generated by the CT data processing module, and perform the thermal-mechanical coupled simulation of steps S213 and S214, based on a user-input material property library and welding process parameters (as described in steps S211 and S212). Its output is a large data file containing displacement vectors of all nodes at all time steps, i.e. a predicted thermal-induced distortion field map.

[0126] A distortion-compensated trajectory generation module, which can be deployed on the same computer as the first two modules. This module is configured to receive both the nominal weld trajectory and the predicted thermal distortion field map as inputs, and perform the calculations of steps S311-S314. It generates a distortion-compensated dynamic welding trajectory program that can be directly downloaded to the robot controller, through spatio-temporal interpolation, vector superposition, and robot pose solving.

[0127] A real-time geometry sensing module, which is physically mounted on the robot end effector. The core of this module is a laser profilometry sensor (consisting of a line laser emitter and a high-speed camera) and its accompanying image acquisition card. This module is configured to trigger and acquire the profile image in front of the weld at a frequency of 200 Hz during the welding process according to the synchronization signal of the robot controller, and perform the image processing and three-dimensional reconstruction algorithms of steps S411-S414. Its output is the current actual weld center position Z_k in the robot base coordinate system at each sampling time k.

[0128] A state estimation and control module, which is usually running on the robot controller or an embedded industrial PC in high-speed communication with it. The core of this module is the Kalman filter state observer described in step S5. It is configured to receive the predicted position from the distortion-compensated trajectory generation module and the measured position Z_k from the real-time geometry sensing module at each sampling period. By performing the filtering algorithm of steps S511-S515, this module calculates the optimal deviation vector and outputs it as a real-time correction signal to the underlying servo controller of the robot, thereby closing the entire high-precision centering control loop.

[0129] The above, only the preferred embodiment of the present application, not to the other forms of the present application limit, any skilled in the art may use the above disclosed technical content to change or modify the equivalent variation of the equivalent embodiments applied to other fields, but any simple modification, equivalent variation and modification of the above embodiments, which does not deviate from the technical solution of the present application, according to the technical essence of the present application, still belongs to the protection scope of the technical solution of the present application.

Claims

1. A method for automatic centering of a light beam based on industrial CT, characterized in that, The method comprises the following steps: S1, obtaining three-dimensional industrial CT volume data of a workpiece to be welded, performing image segmentation and three-dimensional reconstruction processing on the three-dimensional industrial CT volume data, extracting a three-dimensional point cloud model of a weld area, and generating a nominal weld trajectory line representing a weld geometry based on the three-dimensional point cloud model; S2, establishing a thermal-mechanical coupling finite element analysis model matched with material properties and welding process parameters of the workpiece to be welded, performing numerical simulation of a welding process by taking the nominal weld trajectory line as a heat source movement path, and generating a predicted thermal distortion field atlas; S3, superimposing a displacement vector corresponding to the nominal weld trajectory line in a time sequence in the predicted thermal distortion field atlas to the nominal weld trajectory line, and generating a dynamic welding trajectory compensated by thermal distortion; The generation step of the dynamic welding trajectory is specifically: S311, representing the nominal weld trajectory line as a function with arc length as a parameter; S312, querying and obtaining a predicted three-dimensional displacement vector corresponding to a position of the nominal weld trajectory line when welding is performed to a specific arc length from the predicted thermal distortion field atlas; S313, performing vector addition of the predicted three-dimensional displacement vector and a corresponding point coordinate on the nominal weld trajectory line, and obtaining the dynamic welding trajectory; S314, discretizing the dynamic welding trajectory into a series of six-degree-of-freedom path points containing position and attitude information, and forming an instruction sequence for execution by a welding robot controller; S4, in the welding process, collecting real-time three-dimensional profile data of a weld area in front of a welding beam action point by a laser profile measurement sensor fixedly connected with the welding beam, and extracting a current actual weld center position from the real-time three-dimensional profile data; The extraction step of the current actual weld center position is specifically: S411, collecting a laser line image reflecting a weld cross section profile by a line laser emitter and a high-speed industrial camera; S412, performing denoising and binarization processing on the collected image, and extracting center pixel coordinates of the laser line by using a barycenter method or a Steger algorithm, and generating a two-dimensional point set; S413, converting the two-dimensional point set from a pixel coordinate system to a three-dimensional space coordinate system according to a pre-calibrated camera intrinsic matrix and a laser plane equation, and obtaining the real-time three-dimensional profile data; S414, performing V-shaped groove slope line intersection point or profile lowest point searching on the real-time three-dimensional profile data, and determining the current actual weld center position.

2. The industrial CT based light beam automatic centering method according to claim 1, wherein, The three-dimensional point cloud model comprises a vertex coordinate set and a triangular mesh topology, the predicted thermal distortion field atlas comprises finite element model node identification, time step and corresponding three-dimensional displacement vectors, the dynamic welding trajectory comprises a six-degree-of-freedom path point sequence sorted by time stamp and containing position and attitude information, and the real-time three-dimensional profile data comprises a group of three-dimensional space point coordinates representing a weld cross section.

3. The industrial CT based light beam automatic centering method according to claim 1, wherein, The generation step of the nominal weld trajectory line is specifically: S111, performing processing on the three-dimensional industrial CT volume data by using a threshold-based region growing algorithm or a deep learning semantic segmentation network, identifying and segmenting a weld gap and a groove area, and forming independent weld geometry data; S112, apply a three-dimensional body surface extraction algorithm to the weld geometry data to generate the three-dimensional point cloud model representing the inner and outer surfaces of the weld; S113, slice the three-dimensional point cloud model along the predetermined welding direction to obtain a series of two-dimensional weld cross-section profiles, and use geometric center calculation or bottom V-shaped tip recognition algorithm to determine the center point coordinates of each section; S114, connect a series of center point coordinates in three-dimensional space, and use B-spline curve or cubic spline interpolation algorithm for fitting to generate the nominal weld trajectory line.

4. The industrial CT based light beam automatic centering method according to claim 3, characterized in that, The generation step of the predicted thermal distortion field map is specifically: S211, construct a finite element geometric model that reproduces the geometry of the workpiece to be welded, and set the temperature-dependent thermal conductivity, specific heat capacity, density, elastic modulus, Poisson's ratio and thermal expansion coefficient material attribute parameters; S212, equivalent the heat source effect of the welding beam to a three-dimensional Gaussian distribution or double-ellipsoid volume heat source model, and set its power, thermal efficiency and geometric size parameters according to the pre-set welding process; S213, discretize the nominal weld trajectory line, drive the volume heat source model to move along it, and solve the non-steady-state heat conduction equation at each time step to calculate the three-dimensional temperature field distribution of the entire workpiece model; S214, use the three-dimensional temperature field distribution as the thermal load to solve the elastic-plastic constitutive relationship equation based on the thermal strain increment to calculate the displacement vector of all nodes on the workpiece at each time step, forming the predicted thermal distortion field map.

5. The industrial CT based light beam automatic centering method of claim 1, wherein, The method further comprises a step S5: S5, establish a Kalman filter state observer, use the weld center position indicated by the dynamic welding trajectory as the system state prediction value, and use the actual weld center position as the system state observation value, perform optimal state estimation, calculate the deviation vector between the predicted position and the actual position, and generate a real-time position correction control signal for the welding beam actuator; The deviation vector includes lateral position deviation and longitudinal position deviation.

6. The industrial CT based light beam automatic centering method according to claim 5, wherein, The generation step of the real-time position correction control signal is specifically: S511, define a system state vector containing position deviation and deviation change rate; S512, establish a system state prediction model based on a state transition matrix to perform prior state estimation; S513, establish a system observation model based on an observation matrix, and use the difference between the dynamic welding trajectory position and the actual weld center position as the observation value; S514, perform Kalman filter update step, calculate Kalman gain, and update state estimation combined with the observation value to obtain posterior state estimation; S515, extract the position deviation component from the posterior state estimation as the deviation vector, and generate the real-time position correction control signal.

7. An industrial CT-based light beam automatic centering system, characterized in that, The system is used to implement the beam automatic centering method based on industrial CT according to any one of claims 1-6, and the system comprises: a nominal trajectory generation module that acquires three-dimensional industrial CT volume data of a workpiece to be welded, generates a three-dimensional point cloud model through image segmentation and three-dimensional reconstruction, and calculates a nominal weld trajectory line by slicing and center point fitting along the weld direction based on the model; The thermal distortion simulation module constructs a thermal-mechanical coupled finite element model based on the geometry and material properties of the workpiece, simulates the welding process by taking the nominal weld seam trajectory as a heat source path, solves the non-steady-state heat conduction and elastic-plastic mechanics equations, and generates a predicted thermal distortion field atlas; The dynamic trajectory compensation module extracts displacement vectors corresponding to each point on the nominal weld seam trajectory based on the predicted thermal distortion field atlas, performs distortion compensation through vector superposition operation, and generates a dynamic welding trajectory; The real-time contour sensing module drives a laser contour measurement sensor during the welding process, collects real-time three-dimensional contour data in front of the weld seam, and extracts the current actual weld seam center position through image processing and coordinate transformation; The state estimation and correction module establishes a Kalman filter state observer, fuses the predicted position given by the dynamic welding trajectory and the actual position measured by the real-time contour sensing module, performs optimal state estimation, calculates the deviation vector, and generates a real-time position correction control signal for the welding light beam actuator.

Images