Inward Flanging Welding Method for Water Tank Inner Liner and Water Tank Inner Liner

Abstract

The present invention provides an inturned edge welding method for a water tank inner liner and a water tank inner liner, which relates to the technical field of welding manufacturing. The method includes: acquiring three-dimensional point cloud data of the inturned edge interface of the water tank inner liner in real time, and registering the point cloud data with a preset three-dimensional geometric model; planning an initial motion trajectory of a welding torch based on an improved RRT algorithm and in combination with geometric features represented by the registered three-dimensional point cloud data; performing welding based on the initial motion trajectory, and extracting weld width, penetration features and defect features from a molten pool image obtained during the welding process; correcting welding parameters of the welding torch in real time based on the weld width and penetration features; when the defect features do not meet preset conditions, triggering path replanning, and generating a real-time motion trajectory and synchronously adjusted welding parameters based on the currently corrected welding parameters and defect features, and controlling the welding torch to perform compensatory welding. Through the collaborative innovation of multi-modal perception, intelligent trajectory planning and closed-loop parameter control, the present invention improves the welding quality.

Description

Technical Field

[0001] The present invention relates to the technical field of welding manufacturing, and particularly relates to an inturned edge welding method for a water tank inner liner and a water tank inner liner. Background Art

[0002] In the field of water tank inner liner manufacturing, the welding quality of inturned edge joints directly affects the sealing performance, pressure resistance performance and service life of products. The traditional welding process has the following technical defects:

[0003] Poor geometric adaptability: The water tank inner liner mostly adopts a thin-walled curved surface structure, and there are problems such as sudden curvature changes and complex geometric features in the weld transition zone at the inturned edge joint. Conventional automated welding systems rely on offline programming and are difficult to adapt to the real-time changes in the three-dimensional geometric features of the weld, which easily leads to deviations in the torch posture and uneven heat input, and further causes defects such as lack of fusion and cracks.

[0004] Lagged defect response: The dynamic behavior of the molten pool during the welding process is complex. The traditional vision detection system can only extract the macroscopic features of the weld through two-dimensional images and cannot accurately quantify three-dimensional defect information such as penetration depth and pore distribution, resulting in lagged defect compensation compared to the actual working conditions and a low welding qualification rate.

[0005] Severe thermal deformation interference: Welding thermal deformation of the thin-walled structure causes the spatial position of the weld to shift. The traditional Iterative Closest Point (ICP) registration algorithm is sensitive to local geometric distortions caused by thermal deformation, and insufficient registration accuracy directly leads to trajectory planning errors, exacerbating the accumulation of welding defects.

[0006] Therefore, it is necessary to provide an inturned edge welding method for a water tank inner liner and a water tank inner liner to solve the above technical problems. Summary of the Invention

[0007] To solve the above technical problems, the present invention provides an inturned edge welding method for a water tank inner liner and a water tank inner liner, which breaks through the geometric adaptability bottleneck of the traditional welding process and improves the welding quality through the collaborative innovation of multi-modal perception, intelligent trajectory planning and closed-loop parameter control.

[0008] The present invention provides an inturned edge welding method for a water tank inner liner, and the method includes the following steps:

[0009] Obtain the three-dimensional point cloud data of the inturned edge joint of the water tank inner liner in real time, and register the point cloud data with a preset three-dimensional geometric model;

[0010] Based on the improved RRT algorithm, combined with the geometric features represented by the registered three-dimensional point cloud data, plan the initial motion trajectory of the welding torch;

[0011] Perform welding based on the initial motion trajectory, and extract the weld width, penetration depth features and defect features from the molten pool images obtained during the welding process;

[0012] Based on the weld width and penetration characteristics, the welding parameters of the welding torch are corrected in real time;

[0013] When the defect characteristics do not meet the preset conditions, path replanning is triggered, and a real-time motion trajectory and synchronized adjusted welding parameters are generated based on the currently corrected welding parameters and the defect characteristics, and the welding torch is controlled to perform compensated welding.

[0014] Preferably, the real-time acquisition of the three-dimensional point cloud data of the inner flanging interface of the water tank inner liner and the registration of the point cloud data with the preset three-dimensional geometric model include:

[0015] Collect the initial point cloud data of the inner flanging interface at a preset scanning frequency and preprocess the initial point cloud data;

[0016] Extract the edge contour features of the preprocessed initial point cloud data, and register the initial point cloud data with the preset three-dimensional geometric model by the improved ICP method to generate geometric features including the weld edge contour, curvature distribution and normal direction.

[0017] Preferably, the registration by the improved ICP method includes:

[0018] Taking the edge contour features of the preprocessed initial point cloud data as constraint conditions, constructing a feature descriptor including the normal direction angle and curvature gradient, and selecting a weighted corresponding point pair set based on the similarity measure of the feature descriptor;

[0019] In the iterative closest point registration process, the registration error function is dynamically adjusted according to the feature similarity weight of the corresponding point pairs, where the feature similarity weight is determined by a joint function composed of the normal direction angle and curvature gradient;

[0020] Taking the dynamically adjusted registration error function as the input, the registration deviation caused by welding thermal deformation is corrected by the thin plate spline interpolation algorithm.

[0021] Preferably, the initial motion trajectory of the welding torch is planned based on the improved RRT algorithm and the geometric features characterized by the registered three-dimensional point cloud data, including:

[0022] In the path extension process of the improved RRT algorithm, the search weight is dynamically adjusted according to the curvature distribution. When the curvature exceeds the preset threshold, the search weight is reduced by a preset ratio;

[0023] Based on the normal direction and the preset equipment anti-collision constraint, a composite constraint condition including the reachable space model and the obstacle distance field is constructed, and under the limitation of the composite constraint condition, a candidate path set is generated by restricted random sampling;

[0024] Verify the smoothness of all candidate paths by presetting the weld continuity constraint conditions, and select the candidate paths that pass the preset heat input uniformity constraint verification from the candidate paths that pass the smoothness verification as the initial motion trajectory.

[0025] Preferably, weld based on the initial motion trajectory, and extract the weld width, penetration characteristics, and defect characteristics from the molten pool images obtained during the welding process, including:

[0026] Real-time collect the multi-spectral images of the molten pool in the welding area of the flanging interface;

[0027] Calculate the weld width, penetration characteristics, and defect characteristics through the multi-spectral images of the molten pool, where the defect characteristics include three defect types: pores, lack of fusion, and cracks.

[0028] Preferably, based on the weld width and penetration characteristics, real-time correct the welding parameters of the welding torch, including:

[0029] Compare the weld width and penetration characteristics extracted in real-time with the preset target values, and calculate the width deviation and depth deviation;

[0030] Correct the welding parameters according to the width deviation and depth deviation;

[0031] Verify the corrected welding parameters based on the preset heat input uniformity constraint until the verified welding parameters are obtained.

[0032] Preferably, the compensated welding includes:

[0033] Determine the defect area for compensated welding according to the positions and sizes of the defect types in the defect characteristics that do not meet any of the following preset conditions, where the preset conditions are:

[0034] The diameter of the pores ≤ 0.5 mm, and the distribution density ≤ 3 pieces / cm 2 ;

[0035] The length of the lack of fusion ≤ 1 mm, and the depth ≤ 0.3 mm;

[0036] The length of the crack ≤ 0.3 mm, and the orientation angle deviates from the weld center line ≤ 15°;

[0037] Based on the currently corrected welding parameters and the normal direction of the defect area, generate a compensation trajectory that meets the composite constraint conditions through an improved RRT algorithm;

[0038] Dynamically adjust the welding parameters according to the curvature distribution characteristics of the compensation trajectory and verify;

[0039] Control the welding torch to perform the welding operation according to the compensation trajectory and the verified welding parameters.

[0040] The present invention also provides a water tank inner liner, and the water tank inner liner includes an inward flanging interface formed by welding through the above method.

[0041] Compared with the related art, the inward flanging welding method and the water tank inner liner provided by the present invention have the following beneficial effects:

[0042] The present invention proposes a method for constructing a feature descriptor based on edge contour constraints, dynamically adjusts the registration error function through a joint function of the normal direction angle and the curvature gradient, corrects the registration deviation caused by thermal deformation by combining the thin plate spline interpolation algorithm, and realizes the registration of weld geometric features.

[0043] Meanwhile, a curvature-sensitive search weight dynamic adjustment mechanism is introduced to construct a composite constraint model that integrates the normal direction, anti-collision constraints, and heat input uniformity, and generates an initial trajectory that meets weld continuity and equipment accessibility. In the compensation welding stage, the local path is replanned by improving the RRT* algorithm, and a dynamic obstacle avoidance strategy and a heat affected zone overlap degree constraint are incorporated to ensure the process adaptability of the compensation trajectory.

[0044] In addition, three-dimensional defect features such as weld width, penetration depth, and pores / unfused / cracks are extracted based on the multi-spectral image of the molten pool, and a coupling mapping model between the width deviation, depth deviation, and welding parameters is established. When the defect features exceed the preset threshold, the welding parameters are dynamically adjusted through the parameter coupling model, and the effectiveness of the parameters is verified based on the heat input uniformity constraint, realizing the full-process closed-loop control of welding quality. Brief Description of the Drawings

[0045] Figure 1 It is a schematic flow chart of an inward flanging welding method for a water tank inner liner provided by the present invention.

[0046] Figure 2 It is a schematic structural diagram of the water tank inner liner of the present invention. Detailed Embodiments

[0047] The present invention will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. Additionally, it should be noted that for the sake of description, only parts related to the present invention rather than all structures are shown in the drawings. Moreover, without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0048] It should also be noted that, for ease of description, only the parts related to the present invention rather than all the content are shown in the drawings. Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the operations can be rearranged. The process can be terminated when its operations are completed, but it can also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, and so on.

[0049] Embodiment 1

[0050] As the core component of a pressure-bearing container (such as a water heater), the inner flanging welding of the water tank inner liner is a key process in the field of container manufacturing. Specifically: a flange structure that folds inward is processed on the edge of the opening of the shell to form an inner flanging interface. The typical parameters are a flanging height H = 3 - 5 mm and a bending radius R = 0.5 - 1.5t (t is the thickness of the sheet).

[0051] However, the traditional welding process has the following defects:

[0052] During the flanging forming process of thin-walled sheets, springback deformation is likely to occur, resulting in a deviation of 0.3 - 0.8 mm between the actual weld and the theoretical design, and it is difficult to match the traditional off-line programming trajectory.

[0053] The welding heat input causes local shrinkage in the inner flanging area, resulting in dynamic offset of the weld spatial position and further exacerbating the trajectory tracking error.

[0054] Defects such as pores, lack of fusion, and cracks are likely to expand into crack sources in the complex curved surface stress field, reducing the service life of the product.

[0055] Therefore, the present invention provides an inner flanging welding method for the water tank inner liner to solve the above technical problems.

[0056] Specifically, the method of the present invention includes the following steps:

[0057] S1: Real-time obtain the three-dimensional point cloud data of the inner flanging interface of the water tank inner liner, and register the point cloud data with a preset three-dimensional geometric model.

[0058] Specifically, step S1 includes the following steps:

[0059] S11: Collect the initial point cloud data of the inner flanging interface at a preset scanning frequency, and preprocess the initial point cloud data.

[0060] In this embodiment, an initial point cloud data of the flanging interface is acquired by using, including but not limited to, a 3D line laser scanner at a scanning frequency of 500 Hz. The specific implementation is as follows:

[0061] Scanning parameter setting: Laser wavelength: 650 nm (red visible light), scanning line spacing , Z-axis resolution ;

[0062] Then, data preprocessing is performed on the acquired initial point cloud data, which includes:

[0063] Outlier removal: Based on statistical filtering, remove points exceeding the mean range, where is the standard deviation of the point cloud density (calculation formula: , where, represents the density of the point neighborhood, represents the average density, represents the total number of points).

[0064] Curvature non-uniform filtering: Set the curvature threshold , and retain points with a curvature greater than (curvature calculation: , where , are the minimum and maximum eigenvalues of the point cloud covariance matrix) Multi-scale smoothing: Adopt a three-level moving least squares (MLS) filtering with radii , , to eliminate surface ripples.

[0065] S12: Extract the edge contour features of the preprocessed initial point cloud data, and register the initial point cloud data with a preset three-dimensional geometric model by an improved ICP method to generate geometric features including weld edge contours, curvature distributions, and normal directions.

[0066] In this embodiment, point cloud registration is achieved by an improved ICP algorithm. The specific process is as follows:

[0067] First, extract edge contour features, which specifically include:

[0068] Perform region growing segmentation based on normal vectors on the preprocessed initial point cloud, and extract the curvature continuous region as a weld edge candidate.

[0069] Fit a cylindrical surface (radius tolerance ±0.1 mm) through the RANSAC algorithm, and screen the edge point set that conforms to the flanging geometric features.

[0070] Then, construct feature descriptors, which specifically include:

[0071] For each point Construct the feature vector:

[0072]

[0073] in, Represents the point index in the current point cloud to be registered (i.e. the initial point cloud after preprocessing), Represents the corresponding point index in the preset 3D geometric model (determined by nearest neighbor search).

[0074] It is the angle between the current normal vector and the normal line of the corresponding point of the 3D geometric model, which is used to measure the consistency of the local geometric direction of the point cloud. If the normal vector angle is too large (for example, >30°), it indicates that there are significant geometric differences in the area (such as weld edges or deformation areas), and a higher weight should be given in the registration.

[0075] , the curvature gradient of the current point The curvature gradient of the point corresponding to the model The Euclidean distance of the surface can quantify the degree of surface deformation. The greater the difference in curvature gradient (such as stamping springback or thermal deformation area), the more drastic the local geometric change is, and the registration priority needs to be adjusted through dynamic weights.

[0076] is the point density of the current point, defined as The number of points in a sphere with a radius of 3 mm and centered at is used to evaluate data reliability. Data confidence is high in high-density areas (such as flat areas), and weights need to be reduced in low-density areas (such as edges or noise areas) to avoid mismatching.

[0077] Next, the dynamic error function is optimized, including:

[0078] The feature weight factor is introduced in the ICP iteration, and the optimization objective function is:

[0079]

[0080] in, represents a 3×3 rotation matrix, which represents the spatial rotation transformation of the point cloud. represents a 3×1 translation vector, Indicates the total number of corresponding point pairs finally selected, Represents the first Point coordinates, Indicates the point The new coordinates after applying the rotation transformation R, Indicates that the point Rotate first Translate again to align it with the corresponding point in the target point cloud . represents the point coordinates corresponding to in the preset 3D geometric model point cloud, represents the Euclidean distance operator, represents the value of the registration error function.

[0081] represents the feature similarity weight of the th point, and the specific calculation is carried out according to the following formula:

[0082]

[0083] where, represents the angle between the current normal vector and the normal of the corresponding point in the 3D geometric model, represents the difference in curvature gradient of the current matching point pair (represented by the Euclidean distance), represents the point density of the current point, represents the preset reference density of the corresponding area in the model point cloud.

[0084] Finally, after optimizing the dynamic error function, precise compensation for the welding thermal deformation is carried out through the thin plate spline interpolation algorithm.

[0085] During specific implementation, first, no less than 50 control points are selected at a spacing of 2 mm along the weld path as the deformation reference. By establishing a non-rigid deformation field model, the local shrinkage characteristics caused by thermal deformation can be accurately described. During the compensation process, the algorithm will comprehensively analyze the 3D residual vector generated after dynamic registration. When it detects that the maximum residual exceeds 0.2 mm, the compensation mechanism is automatically triggered.

[0086] The compensation execution uses the least squares method to calculate the optimal correction parameters, including the linear transformation matrix and the translation vector, to ensure that the path deviation after compensation is controlled within the process requirement of ±0.1 mm tolerance range.

[0087] After registration, geometric features are generated, specifically including:

[0088] First, calculate the normal vector of the registered point cloud (neighborhood radius 3 mm), identify the continuous point set with a normal vector angle greater than 30° as the initial weld edge, and then optimize the contour through RANSAC cylindrical surface fitting (radius tolerance ±0.1 mm) to finally obtain a weld edge point set with an accuracy of ±0.03 mm.

[0089] The curvature distribution feature calculates the local surface Gaussian curvature through the moving least squares method, divides the point cloud into 0.5 mm × 0.5 mm × 0.2 mm voxel grids and calculates the mean curvature of each grid to generate a gradient field that can identify curvature mutations above 0.3 mm.

[0090] The initial value of the normal direction is estimated by the PCA method, and the neighborhood radius is reduced to 1 mm in the high-curvature region ( ) to improve the accuracy. Then, the normal orientation is unified by the minimum spanning tree algorithm, and finally, a unit normal vector field with an angular dispersion less than 0.8° is obtained.

[0091] S2: Based on the improved RRT algorithm, combined with the geometric features represented by the registered three-dimensional point cloud data, plan the initial motion trajectory of the welding torch.

[0092] Specifically, step S2 includes the following steps:

[0093] S21: During the path extension process of the improved RRT algorithm, dynamically adjust the search weight according to the curvature distribution. When the curvature exceeds the preset threshold, reduce the search weight by a preset ratio.

[0094] In this embodiment, during the path extension process of the improved RRT algorithm, dynamically adjust the search weight according to the curvature distribution of the registered point cloud to optimize the path planning efficiency and accuracy. The curvature threshold is set to , which is calibrated based on the stamping springback test results of common materials such as 304 stainless steel and aluminum alloy, and can effectively distinguish flat areas from high-curvature features (such as the starting and ending points of welds or bending areas).

[0095] When it is detected that the curvature of the current path extension area exceeds the threshold, the algorithm reduces the search weight of this area by a preset ratio. Exemplarily, when the curvature is , the weight is reduced to 70% of the original value, and when the curvature is , the weight is reduced to 50%. The weight adjustment uses linear interpolation to ensure smooth transition. At the same time, in the high-curvature region (curvature > ), the path extension step size is reduced from the default 1 mm to 0.5 mm to avoid path mutations caused by too large a step size. The search direction preferentially expands along the direction of the curvature gradient descent to reduce redundant nodes of the path in complex geometric regions and improve the planning efficiency.

[0096] S22: Based on the normal direction and the preset equipment anti-collision constraints, construct a composite constraint condition including the reachability space model and the obstacle distance field, and generate a candidate path set through restricted random sampling under the restriction of the composite constraint condition.

[0097] In this embodiment, based on the registered normal direction and the equipment kinematic parameters, construct a candidate path set under the composite constraint condition.

[0098] First, define the reachable space according to the kinematic model of the end effector of the welding torch (Denavit-Hartenberg parameters of a six-degree-of-freedom robotic arm), restrict the joint rotation angle range of the welding torch attitude (such as the maximum rotation angle of each axis ±180°) and the working range of the tool center point (TCP) (diameter ≤ 500 mm).

[0099] Combined with the normal direction of the point cloud, force the angle between the axis of the welding torch and the surface normal to be ≤ 5° to prevent the lack of fusion defect caused by wire deviation. Secondly, convert the three-dimensional models of the water tank inner liner and the fixture into an Euclidean distance field, and calculate the minimum safety distance (≥ 3 mm) between the welding torch and the obstacle in real time. This distance is comprehensively set based on the diameter of the end of the welding torch (10 mm) and the thermal expansion margin (0.2 mm).

[0100] During the path extension process, adopt a biased sampling strategy guided by a probability threshold: expand towards the target point with an 80% probability and randomly explore with a 20% probability. At the same time, introduce a repulsive force field in the collision risk area (distance from the obstacle < 5 mm) to dynamically adjust the sampling direction. The candidate path needs to pass the reachability verification and collision avoidance detection to ensure that the node spacing is uniform and there is no interference.

[0101] S23: Verify the smoothness of all candidate paths through the preset weld continuity constraint conditions, and select the candidate paths that pass the preset heat input uniformity constraint verification from the candidate paths that pass the smoothness verification as the initial motion trajectory.

[0102] In this embodiment, perform multi-level verification and optimization on the generated candidate path set to ensure that the welding process requirements are met.

[0103] First, smooth the path through cubic B-spline interpolation, control the curvature change rate between nodes ≤ 0.15 rad / mm, and eliminate attitude mutations. The interpolation parameters are set as follows: the control point spacing is 2 mm, and the curve tension coefficient is 0.5 to ensure path continuity and the smoothness of the welding torch movement.

[0104] Secondly, based on the welding current - speed - heat input model (heat input = current × voltage / welding speed), calculate the heat input of each path segment, and require that the coefficient of variation of the heat input of the entire path ≤ 5% (calculation formula: standard deviation / mean × 100%). For the paths that exceed the standard (such as the coefficient of variation > 5%), automatically adjust the welding speed or current parameters and re-verify after adjustment.

[0105] Finally, select the comprehensive optimal solution from the paths that pass the verification: use the total path length (weight 60%) and heat input uniformity (weight 40%) as evaluation indicators, and use the weighted scoring method (total score = 0.6 × length normalization value + 0.4 × heat input uniformity normalization value) for sorting, and select the path with the highest total score as the initial motion trajectory.

[0106] S3: Perform welding based on the initial motion trajectory, and extract the weld width, penetration features, and defect features from the molten pool images obtained during the welding process.

[0107] Specifically, step S3 includes the following steps:

[0108] S31: Real-time collect the multi-spectral images of the molten pool in the welding area of the inward flanging interface.

[0109] During the welding process, use high-precision multi-spectral imaging technology to real-time collect the dynamic image data of the molten pool. Configure an industrial-grade camera, equipped with a high-performance sensor, supporting synchronous imaging in the visible light (500 - 700nm) and near-infrared (850 - 1050nm) dual channels. The camera is equipped with a standard focal length lens, the aperture is set to f / 2.8, the working distance is fixed at 200mm, and the depth of field range is ±1.5mm to ensure clear imaging of the molten pool area. A band-pass filter is installed in the visible light channel, and a special filter is installed in the near-infrared channel. The attenuation rate of the filter is > 99.5%, effectively suppressing the interference of the welding arc light. The imaging resolution is set to 0.02mm / pixel (corresponding to 2048×1536 pixels), covering a 20mm×15mm welding area (including the molten pool and a 5mm heat-affected zone on each side).

[0110] Send a trigger signal through the bus of the welding torch motion controller, triggering an exposure every 0.5mm of the welding torch displacement to achieve continuous spatial sampling. Dynamically adjust the exposure time (0.1ms - 1ms) in HDR mode to adapt to the drastic change in the brightness of the molten pool.

[0111] In addition, to ensure data reliability, use optical fiber to transmit the image data to shield electromagnetic interference, and install an air curtain device (flow rate 5L / min) to blow away the welding fumes in real-time to ensure image clarity.

[0112] S32: Calculate the weld width, penetration features, and defect features from the multi-spectral images of the molten pool, where the defect features include three defect types: pores, lack of fusion, and cracks.

[0113] In this embodiment, when extracting the welding quality features based on the multi-spectral images, a multi-level processing flow is executed.

[0114] For weld width measurement, first perform median filtering (3×3 kernel) on the visible light image to reduce noise, and then use an adaptive threshold segmentation algorithm to extract the binary contour of the molten pool. Set measurement lines every 0.2mm along the normal direction of the weld (based on the normal field data generated in step S12), and calculate the maximum value of the contour edge spacing as the weld width. The measurement accuracy is calibrated by a standard gauge to ensure an error ≤ ±0.05mm (95% confidence level).

[0115] The penetration depth feature is inverted through the near-infrared band: Based on the optical law, an intensity-depth model is established, and the formula is expressed as the penetration depth h equals the reciprocal of the material absorption coefficient multiplied by the natural logarithm of the ratio of the reference light intensity to the measured light intensity . Among them, the calibration value of stainless steel is . The penetration depth curve is generated every 0.2 mm along the weld, and the area where the detected depth mutation > 0.3 mm is used as the lack of fusion defect.

[0116] Defect detection adopts a multi-algorithm fusion strategy: For pores, the closed area is extracted through morphological processing, and the targets with a diameter > 0.5 mm and roundness > 0.7 are screened; For lack of fusion, it is judged by combining the visible light edge gradient and the near-infrared depth mutation; For cracks, the linear features with a length > 0.3 mm, aspect ratio > 5:1, and deviation from the weld center line > 15° are extracted through the edge detection algorithm.

[0117] S4: Based on the weld width and penetration depth features, the welding parameters of the welding torch are corrected in real time.

[0118] Specifically, step S4 includes the following steps:

[0119] S41: Compare the weld width and penetration depth features extracted in real time with the preset target values, and calculate the width deviation and depth deviation.

[0120] During the real-time welding process, the collected weld width and penetration depth features are dynamically compared with the preset target values. The preset target values are set according to the material thickness and process specifications. For example, for the 304 stainless steel inner liner with a thickness of 2 mm, the target weld width is 2.0 ± 0.1 mm, and the target penetration depth is 1.5 ± 0.05 mm. The real-time data is extracted every 0.2 mm of the welding torch displacement, and the width deviation and depth deviation are calculated: The width deviation is the absolute difference between the measured value and the target value, and the depth deviation is also the absolute difference between the measured value and the target value. The deviation data is processed by sliding window filtering (window length 10 mm), and the smoothed width deviation and depth deviation are output after eliminating local noise interference.

[0121] S42: Correct the welding parameters according to the width deviation and depth deviation.

[0122] In this embodiment, when dynamically adjusting the welding parameters according to the deviation, a multi-variable coupling control strategy is adopted. When the width deviation > 0.1 mm, the welding current is adjusted proportionally: For every increase of 0.1 mm, the current is reduced by 5 A (for example, from 150 A to 145 A); For every 0.1 mm decrease, the current increases by 5 A.

[0123] If the penetration deviation > 0.05 mm, synchronously adjust the current: For every 0.05 mm increase, the current increases by 3 A to compensate for insufficient penetration; For every 0.05 mm decrease, the current is reduced by 3 A to avoid excessive penetration. When and both exceed the standard, give priority to adjusting the welding speed: the speed adjustment amount = 0.2 × ΔW + 0.3 × ΔD (unit: mm / s). For example = 0.15 mm, = 0.06 mm, = 0.048 mm / s (the speed is adjusted from 5 mm / s to 5.048 mm / s). The voltage is synchronously fine-tuned based on the preset current-voltage characteristic curve to maintain arc stability, and the adjustment range ≤ ±2 V.

[0124] S43: Verify and correct the welding parameters based on the preset heat input uniformity constraint until the verified welding parameters are obtained.

[0125] In this embodiment, the corrected parameters need to pass the heat input uniformity verification. Calculate the heat input per unit length Q = (current I × voltage U) / speed V (unit: J / mm) in real time. Divide the weld into analysis segments every 5 mm, and count the average value of the Q values in each segment and the standard deviation , calculate the coefficient of variation C = / × 100%. If C > 5%, trigger secondary correction: for the section where the Q value fluctuates > 10%, adjust the speed (preferably) or current proportionally, and the adjustment range ≤ ±5%. After correction, collect data again for verification until C of three consecutive analysis segments is ≤ 5%. After passing the verification, lock the parameters and record the historical data. If the correction fails to meet the standard for 5 consecutive times, suspend welding and alarm. After testing, this method can control the coefficient of variation of heat input within 4.2%, and the fluctuations of weld width and penetration are respectively ≤ ±0.07 mm and ±0.03 mm, meeting the process requirements of the water tank inner liner.

[0126] S5: When the defect characteristics do not meet the preset conditions, trigger path replanning, and generate a real-time motion trajectory and synchronously adjusted welding parameters based on the currently corrected welding parameters and the defect characteristics, and control the welding torch to perform compensation welding.

[0127] Specifically, the compensation welding in step S5 includes:

[0128] S51: Determine the defective area for compensating welding based on the positions and dimensions of the defective types that do not meet any of the following preset conditions, where the preset conditions are:

[0129] The diameter of the porosity ≤ 0.5 mm, and the distribution density ≤ 3 pieces / cm 2 ;

[0130] The length of the lack of fusion ≤ 1 mm, and the depth ≤ 0.3 mm;

[0131] The length of the crack ≤ 0.3 mm, and the orientation angle deviates from the weld center line ≤ 15°.

[0132] In this embodiment, welding defects are detected in real time through multi-spectral image analysis. When the diameter of the porosity exceeds 0.5 mm or the distribution density is greater than 3 pieces / cm², its geometric center is determined through morphological processing and a compensation area with a radius of 0.75 mm is delimited; for lack of fusion defects, combining the melt depth curve and edge gradient detection, when the melt depth difference of three consecutive points exceeds 0.3 mm and the length is greater than 1 mm, it is extended by 0.5 mm along the weld direction as the compensation area; for crack defects, linear features with a length exceeding 0.3 mm and deviating from the weld center line by more than 15° are screened through edge detection, and a compensation area is delimited by extending 0.2 mm at both ends. All defective areas are extended outward by 0.3 mm as the anti-leakage welding allowance. When the distance between adjacent defects is less than 0.5 mm, they are automatically merged. Finally, the image coordinates are converted to the three-dimensional point cloud space to generate a compensation area model with normal vectors.

[0133] S52: Based on the currently corrected welding parameters and the normal direction of the defective area, generate a compensation trajectory that meets the composite constraint conditions through an improved RRT algorithm.

[0134] In this embodiment, based on the normal direction of the defective area and the current welding parameters, a compensation trajectory is generated through an improved RRT algorithm. First, load the composite constraint conditions (reachable space model + obstacle distance field) of step S2, and start local path planning in the defective area:

[0135] Search for the starting point: The geometric center of the defective area is offset by 0.5 mm along the normal direction (avoiding the welded area).

[0136] Target point: Extend 1 mm along the weld direction at the end of the defect.

[0137] Step size adjustment: The step size in the defective area is set to 0.3 mm (30% of the default step size).

[0138] Weight optimization: The weight of the area where the curvature > is reduced to 40%, and the defect-dense area is preferentially bypassed

[0139] During the planning process, the distance between the welding torch and the obstacle is detected in real time (≥3 mm), and candidate paths that meet the anti-collision constraints are generated. Finally, the path with a thermal input coefficient of variation <3% and the shortest length is selected as the compensation trajectory.

[0140] S53: Dynamically adjust the welding parameters according to the curvature distribution characteristics of the compensation trajectory and verify.

[0141] In this embodiment, the welding parameters are dynamically adjusted according to the geometric characteristics of the compensation trajectory:

[0142] Curvature self-adaptive adjustment:

[0143] When the curvature > the welding speed is reduced by 20% (e.g., from 5 mm / s to 4 mm / s).

[0144] For every increase in curvature of the current is increased by 3 A (e.g., from 150 A to 153 A).

[0145] Normal direction calibration:

[0146] When the angle between the axis of the welding torch and the normal line > 3°, attitude fine-tuning is triggered, with a maximum adjustment angular velocity of 0.5 rad / s.

[0147] Thermal input verification:

[0148] Calculate the thermal input Q of the compensation section = (current × voltage) / speed, and require the Q value to fluctuate ≤ ±5%.

[0149] In the over-standard area (Q fluctuation > 5%), the current is automatically reduced by 2 A or the speed is increased by 0.1 mm / s.

[0150] After the parameters are adjusted, the feasibility of the trajectory is verified through virtual welding simulation. If a collision risk or over-standard thermal input is detected, re-planning is performed.

[0151] S54: Control the welding torch to perform welding operations according to the compensation trajectory and the verified welding parameters.

[0152] In this embodiment, when controlling the welding torch to perform compensation welding, the quintic polynomial interpolation algorithm is used to plan the movement trajectory of the welding torch to ensure that the acceleration ≤ 0.2g to eliminate mechanical vibration.

[0153] During the compensation welding process, multi-spectral images of the molten pool are collected in real time for every 0.1 mm displacement, and the weld width (allowable error ±0.05 mm) and penetration depth (allowable error ±0.03 mm) are verified synchronously.

[0154] Set triple termination conditions:

[0155] 1) Confirm through image processing that the coverage rate of the defect area > 95%;

[0156] 2) The coefficient of variation of the heat input in the compensation section C ≤ 3%;

[0157] 3) There is no new defect alarm (porosity / incomplete fusion / crack) during real-time monitoring.

[0158] The welding parameters are dynamically adjusted according to the geometric characteristics of the compensation area: in the area where the curvature > the welding speed is automatically reduced by 20% and the current is increased by 3 A, while keeping the angle between the axis of the welding torch and the normal line ≤ 3°. If the compensation fails to meet the standard for 3 consecutive times (any condition is not satisfied), the operation is immediately paused and a maintenance alarm is triggered. At the same time, the current welding parameters, trajectory coordinates, and molten pool image data are recorded for fault analysis.

[0159] Embodiment 2

[0160] The present invention also provides a water tank inner liner, and the water tank inner liner 1 includes an inward flanging interface 2 formed by welding through the above method.

[0161] This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for realizing the functions specified in one block or multiple blocks.

[0162] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing relevant hardware through a program, and this program can be stored in a computer-readable storage medium. The storage medium includes read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc memories, magnetic disk memories, tape memories, or any other medium that can be used to carry or store data and is computer-readable.

[0163] It should also be noted that the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or also includes elements inherent in such a process, method, commodity or device. Without further limitation, an element defined by the statement "including one..." does not exclude the existence of another identical element in the process, method, commodity or device including the element.

Claims

1. A method for inward flanging welding of a water tank inner liner, characterized in that, The method includes the following steps: Obtain the three-dimensional point cloud data of the inner flanging interface of the water tank inner liner in real time, and register the point cloud data with a preset three-dimensional geometric model. This process specifically includes: Collect the initial point cloud data of the inner flanging interface at a preset scanning frequency, and preprocess the initial point cloud data; Extract the edge contour features of the preprocessed initial point cloud data, and register the initial point cloud data with the preset three-dimensional geometric model by an improved ICP method to generate geometric features including the weld edge contour, curvature distribution, and normal direction, Among them, the registration by the improved ICP method includes: Construct a feature descriptor including the normal direction angle and curvature gradient with the edge contour features of the preprocessed initial point cloud data as constraint conditions, and select a weighted corresponding point pair set based on the similarity measure of the feature descriptor; During the iterative closest point registration process, dynamically adjust the registration error function according to the feature similarity weight of the corresponding point pair, where the feature similarity weight is determined by a joint function composed of the normal direction angle and curvature gradient; Take the dynamically adjusted registration error function as the input, and correct the registration deviation caused by welding thermal deformation through the thin plate spline interpolation algorithm; Based on the improved RRT algorithm, combined with the geometric features represented by the registered three-dimensional point cloud data, plan the initial motion trajectory of the welding torch. Specifically, the planning of the initial motion trajectory includes: During the path extension process of the improved RRT algorithm, dynamically adjust the search weight according to the curvature distribution. When the curvature exceeds the preset threshold, reduce the search weight by a preset ratio; Construct a composite constraint condition including an accessibility space model and an obstacle distance field based on the normal direction and the preset equipment anti-collision constraint, and generate a candidate path set through restricted random sampling under the limitation of the composite constraint condition; Verify the smoothness of all candidate paths through the preset weld continuity constraint condition, and select the candidate path that passes the verification of the preset heat input uniformity constraint from the candidate paths that pass the smoothness verification as the initial motion trajectory; Perform welding based on the initial motion trajectory, and extract the weld width, penetration depth features, and defect features from the molten pool image obtained during the welding process. Among them, the defect features include three defect types: pores, lack of fusion, and cracks; Based on the weld width and penetration depth features, real-time correct the welding parameters of the welding torch; When the defect features do not meet the preset conditions, trigger path replanning, and generate a real-time motion trajectory and synchronously adjusted welding parameters based on the currently corrected welding parameters and the defect features, and control the welding torch to perform compensation welding. Among them, the compensation welding includes: Determine the defect area of the compensation welding according to the position and size of the defect type that does not meet any of the following preset conditions in the defect features. The preset conditions are: The diameter of the described pores is ≤ 0.5 mm, and the distribution density is ≤ 3 pores / cm 2 ; The length of the lack of fusion ≤ 1 mm and the depth ≤ 0.3 mm; The length of the crack ≤ 0.3 mm and the orientation angle deviates from the weld center line ≤ 15°; Based on the currently corrected welding parameters and the normal direction of the defect area, generate a compensation trajectory that meets the composite constraint conditions through the improved RRT algorithm; Dynamically adjust the welding parameters according to the curvature distribution characteristics of the compensation trajectory and verify them; Control the welding torch to perform welding operations according to the compensation trajectory and the verified welding parameters.

2. The inside flanging welding method of a water tank inner liner according to claim 1, characterized in that, Welding is performed based on the initial motion trajectory, and the weld width, penetration characteristics, and defect characteristics are extracted from the molten pool images obtained during the welding process, including: Real-time collect the multi-spectral images of the molten pool in the welding area of the inward flanging interface; Calculate the weld width, penetration characteristics, and defect characteristics through the multi-spectral images of the molten pool.

3. A method for inturned edge welding of the inner tank of a water tank according to claim 2, characterized in that, Based on the weld width and penetration characteristics, the welding parameters of the welding torch are corrected in real time, including: Compare the weld width and penetration characteristics extracted in real time with the preset target values, and calculate the width deviation and depth deviation; Correct the welding parameters according to the width deviation and depth deviation; Verify the corrected welding parameters based on the preset heat input uniformity constraint until the verified welding parameters are obtained.

4. A water tank inner liner, characterized in that, The water tank inner liner includes an inward flanging interface formed by welding using the method according to any one of claims 1 to 3.

Images