Method for compensating for manufacturing and assembly deviations in the welding of a segmented suction sail for a ship

Abstract

The application provides a segmented manufacturing welding shrinkage and assembly deviation compensation method for a suction sail of a ship, comprising: if a local shrinkage vector in a prediction distribution diagram exceeds a preset threshold, adjusting welding sequence parameters through an iterative optimization algorithm to obtain an optimized welding path planning; extracting segmented boundary coordinates from the optimized welding path planning, performing three-dimensional scanning data fusion on a curved surface structure, and judging a cumulative effect of a size deviation; obtaining the judged cumulative effect of the size deviation, simulating an assembly fitting process by using a grid deformation algorithm, and determining a deviation correction matrix of a potential misalignment area; applying a virtual force field constraint to the segmented model through the determined deviation correction matrix to obtain a corrected segmented geometry shape; generating a digital twin model for assembly alignment according to the corrected segmented geometry shape, judging whether an overall precision meets a preset standard, and obtaining a final manufacturing adjustment scheme.

Description

Technical Field

[0001] This invention relates to the field of shipbuilding technology, and in particular to a method for compensating for welding shrinkage and assembly deviations in the segmented manufacturing of marine suction sails. Background Technology

[0002] The manufacture of large-size marine suction sails is a crucial technological field in the shipbuilding industry, directly impacting ship performance and navigation efficiency, and possessing irreplaceable value, particularly in energy conservation, emission reduction, and shipping economics. As a vital structural component of a ship, the manufacturing precision and quality of the suction sail have a profound influence on overall performance; therefore, controlling dimensional deviations during the manufacturing process has become a focal point of industry attention.

[0003] However, current manufacturing methods in this field still have significant shortcomings. Many traditional processes often focus only on error control in a single stage, neglecting the issue of accumulated deviations throughout the entire process, from material cutting to final assembly. This one-sidedness leads to the inability to effectively connect and comprehensively manage errors at different stages of the manufacturing process. Especially in the assembly of complex curved structures, the cumulative effect of deviations is often underestimated, ultimately affecting the accuracy and stability of the overall structure.

[0004] Against this backdrop, the core technical challenges in manufacturing large-size suction sail segments lie in the shrinkage deformation generated during welding and the control of deviations in subsequent assembly stages. Welding, as a crucial process for connecting curved outer panels, causes unpredictable shrinkage of the material due to high temperatures. This shrinkage not only affects the dimensional accuracy of individual segments but also amplifies deviations when multiple segments are joined, leading to misalignment or gaps at the joint. For example, if the welding shrinkage of a curved outer panel is not accurately predicted during segment manufacturing, it may prevent it from fitting tightly with adjacent segments during assembly, thus affecting the overall structural strength and sealing. This chain reaction from localized shrinkage to overall assembly deviations has become a critical challenge to overcome in the manufacturing process.

[0005] Therefore, how to effectively predict and control the dimensional changes caused by welding shrinkage throughout the entire manufacturing process, and on this basis, achieve precise alignment when assembling the sections, has become a key issue in the manufacturing of large-size marine suction sails. Summary of the Invention

[0006] This invention provides a method for compensating for welding shrinkage and assembly deviations in the segmented manufacturing of marine suction sails, mainly including: A thermodynamic segmented model of the welding process is constructed using the finite difference method. The temperature distribution field is obtained from the preset material properties and welding parameters, and the initial shrinkage vector of each segment is obtained. Based on the obtained initial shrinkage vector, a prediction model is trained using the support vector machine algorithm, and features are extracted from historical welding data to determine the predicted distribution map of shrinkage deformation. If the local shrinkage vector in the predicted distribution map exceeds the preset threshold, the welding sequence parameters are adjusted through an iterative optimization algorithm to obtain an optimized welding path plan. Segment boundary coordinates are extracted from the optimized welding path plan, and three-dimensional scanning data is fused for the curved structure to determine the cumulative effect of dimensional deviations. The cumulative effect of dimensional deviations after judgment is obtained, and the assembly and fitting process is simulated using a mesh deformation algorithm to determine the deviation correction matrix for potential misalignment areas. By applying a virtual force field constraint to the piecewise model using a determined deviation correction matrix, the corrected piecewise geometry is obtained. Based on the corrected segmented geometry, a digital twin model for assembly alignment is generated to determine whether the overall accuracy meets the preset standards, and the final manufacturing adjustment plan is obtained.

[0007] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a method for compensating for welding shrinkage and assembly deviations in the segmented manufacturing of marine suction sails. It proposes a complete solution for complex business scenarios involving shrinkage deformation prediction and assembly accuracy control during welding. The method constructs a thermodynamic segmented model using the finite difference method to obtain the temperature distribution field and initial shrinkage vector. It then trains a prediction model using a support vector machine algorithm to generate a shrinkage deformation distribution map and optimizes welding path planning for local over-threshold regions. Subsequently, the invention integrates 3D scanning data to analyze the cumulative effect of dimensional deviations, uses a mesh deformation algorithm to simulate the assembly and fitting process, determines the deviation correction matrix, applies virtual force field constraints to correct the geometry, and finally generates a digital twin model for assembly alignment, ensuring overall accuracy meets standards. Through multi-stage collaborative optimization, this invention significantly improves the reliability of welding deformation control and assembly accuracy, providing an efficient and precise adjustment solution for the manufacturing of complex curved surface structures. Attached Figure Description

[0008] Figure 1 This is a flowchart of a method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to the present invention.

[0009] Figure 2 This is a schematic diagram of a method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to the present invention.

[0010] Figure 3 This is another schematic diagram of a method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and thoroughly described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of the present invention.

[0012] like Figures 1-3 This embodiment of a method for compensating for welding shrinkage and assembly deviations in the segmented manufacturing of marine suction sails may specifically include: S101. Construct a thermodynamic segmented model of the welding process using the finite difference method, obtain the temperature distribution field from the preset material properties and welding parameters, and obtain the initial shrinkage vector of each segment.

[0013] By using preset material properties and welding parameters, the welding process is meshed using the finite difference method to construct a piecewise thermodynamic model, resulting in the heat conduction equations for each mesh element. Based on these equations, the temperature values ​​of each mesh element at different time steps are iteratively calculated to obtain the dynamic temperature distribution field during the welding process.

[0014] ; Represents grid points , The temperature value at time step n. Indicates the thermal diffusivity. Indicates the time step. This represents the spatial grid spacing. This formula is used to iteratively calculate the temperature distribution field of each grid cell at different time steps.

[0015] For a dynamic temperature distribution field, the temperature gradient changes in each segment are analyzed to determine the correspondence between temperature gradient and thermal stress. Based on this correspondence, the thermal stress distribution in each segment is calculated, deriving preliminary vector data for initial contraction. The finite difference method is then used to correct the boundary conditions of the preliminary vector data, obtaining the accurate initial contraction vector for each segment.

[0016] ; Represents the exact initial contraction vector. This represents the initial contraction vector. Representing the normal derivative, This formula represents the boundary length and is used in the finite difference method to correct the boundary conditions of the initial vector data, thereby obtaining the accurate initial shrinkage vector for each segment.

[0017] If the calculated result of the precise initial shrinkage vector exceeds the preset threshold range, the mesh density is adjusted, the temperature distribution field is recalculated, and the rationality of the final shrinkage vector is judged. Using the adjusted temperature distribution field and shrinkage vector, thermodynamic response data for each segment are generated to determine the overall deformation trend of the welding process.

[0018] A piecewise thermodynamic model of the welding process is constructed using the finite difference method to obtain the temperature distribution field and initial contraction vector. This can be achieved through the following specific implementation method. First, it is assumed that the welding material is low-carbon steel with a thermal conductivity of 50 W / (m·K), a specific heat capacity of 460 J / (kg·K), a density of 7800 kg / m³, a welding heat source power of 2000 W, a welding speed of 0.002 m / s, and an ambient temperature of 25℃. Using the finite difference method, the welding area is divided into a 100×50×20 grid, with each grid cell having a size of 0.001 m. A three-dimensional heat conduction equation is established, and the temperature field is solved using an explicit difference scheme with a time step of 0.01 s. The temperature change at each grid point is calculated iteratively. Considering the influence of heat source movement, the temperature at the center of the heat source can reach 1500℃. The boundary condition is set as convective heat transfer with a heat transfer coefficient of 10 W / (m²·K). During the calculation, the temperature distribution field shows a larger temperature gradient near the heat source and a gradual decrease in temperature at the edges. For example, the temperature is approximately 800℃ at a distance of 0.01m from the heat source, while it drops to 300℃ at 0.05m. Next, based on the temperature field results and the material's coefficient of thermal expansion (set to 1.2e⁻⁵ / ℃), the thermal strain at each grid point is calculated, thus deriving the initial contraction vector. For instance, at grid points near the heat source, the contraction is approximately 0.012m when the temperature changes by 1000℃, while the contraction is only 0.002m further away from the heat source. This method correlates the temperature field with the contraction vector, forming the thermodynamic analysis basis of the piecewise model. To further refine the model, the latent heat of phase change (set to 2.7e⁵ J / kg) can be introduced, and the heat distribution calculation can be adjusted when the temperature reaches the melting point (approximately 1450℃) to ensure model accuracy. All of the above processes can be implemented through programming. For example, using Python in conjunction with the NumPy library to perform matrix operations can automatically complete mesh generation, temperature iteration, and shrinkage vector calculation, ultimately outputting a visualized temperature field and shrinkage distribution map, ensuring the automation and accuracy of data processing.

[0019] S102. Based on the obtained initial shrinkage vector, a prediction model is trained using the support vector machine algorithm, and features are extracted from historical welding data to determine the predicted distribution map of shrinkage deformation.

[0020] A prediction model using the initial shrinkage and vector data is constructed using the Support Vector Machine (SVM) algorithm. Historical data is processed to obtain a structured dataset with extracted features. Based on this structured dataset, the SVM algorithm is used to train the model and determine the preliminary predicted distribution of the shrinkage deformation.

[0021] ;

[0022] This indicates a preliminary predicted distribution. Indicates the number of support vectors. Show the Lagrange multipliers, Indicates historical data tags, Represents the kernel function. This represents a structured dataset sample after feature extraction. Indicates input features, This represents the bias term; this formula is used to determine the preliminary predicted distribution of shrinkage deformation by training the model using the support vector machine algorithm based on the structured dataset after feature extraction.

[0023] For the initial predicted distribution, key time point data during the welding process are acquired to determine the phased changes in the deformation trend. If the fluctuation of the deformation trend at a certain time point exceeds a preset threshold, the historical data is segmented to obtain an adjusted feature dataset. This adjusted feature dataset is reloaded into the prediction model to determine the updated shrinkage deformation distribution map. The updated shrinkage deformation distribution map is then obtained and, combined with the welding process parameters, the stability of the overall deformation trend is assessed. Based on the stability of the overall deformation trend, phased deformation prediction data for the welding process is generated to determine the final distribution prediction result.

[0024] Based on the obtained initial shrinkage vector, a prediction model is constructed and a distribution map of shrinkage deformation is generated. This can be automated using the following method: First, assume the initial shrinkage vector data comes from a 100×100×30 grid within the welding area, with each grid cell's shrinkage value ranging from 0.001m to 0.015m. For example, the shrinkage value of the grid near the welding center is 0.014m, while the shrinkage value of the edge grid is 0.002m. These data serve as input features. Next, a support vector machine algorithm is used for model training. The radial basis function is selected as the kernel function, with a penalty parameter C of 10 and a kernel parameter γ of 0.1. By extracting features from historical welding data, the shrinkage vector is associated with welding parameters (e.g., heat input is set to 1800W, welding speed is 0.003m / s), constructing a training dataset containing 5000 sample points. Each sample point includes the shrinkage value, location coordinates, and process parameters. The algorithm optimizes model parameters through cross-validation (5-fold) to ensure prediction accuracy, with the training set accounting for 80% and the test set for 20%. The final mean squared error of the model on the test set is controlled within 0.0005. Then, the trained model is used to predict new input shrinkage vector data, generating a prediction matrix. For example, the prediction result shows that the shrinkage deformation in the weld center area is 0.013m, and in the edge area it is 0.0015m. The predicted values ​​are smoothed using an interpolation algorithm (such as Kriging interpolation) to generate a continuous shrinkage deformation distribution map, with the grid resolution increased to 200×200×60 to ensure the fineness of the distribution map. The entire process can be implemented using Python combined with the Scikit-learn library for support vector machine training and prediction, automatically completing feature extraction, model optimization, and distribution map generation. Simultaneously, the Matplotlib library is used to save the predicted distribution map as a high-resolution image file, forming a complete data processing chain that ensures logical rigor and requires no manual intervention.

[0025] S103. If the local shrinkage vector in the predicted distribution map exceeds the preset threshold, the welding sequence parameters are adjusted through an iterative optimization algorithm to obtain an optimized welding path plan.

[0026] By analyzing the data from the predicted distribution map, the specific distribution of local contraction vectors is obtained, and the range of areas exceeding a preset threshold is determined. For the range of areas exceeding the preset threshold, an iterative optimization algorithm is used to process the parameter configuration of the welding sequence, resulting in an adjusted parameter set.

[0027]

[0028] This represents the adjusted set of parameters. This indicates the welding sequence parameter configuration. This indicates the range of regions exceeding the threshold. This represents a parameter-based local contraction vector. Represents points within a region.

[0029] Based on the adjusted parameter set, a preliminary optimized path scheme is generated, and its applicability in the distribution analysis is determined. If the path scheme exhibits local mismatches in the distribution analysis, the welding plan is segmented, and segmented path data is obtained. The segmented path data is then reloaded into the iterative optimization algorithm to determine the updated welding sequence scheme. The updated welding sequence scheme is then combined with the distribution analysis results to generate the final welding plan. For the final welding plan, complete optimized path data is output, and the sequence execution logic during the welding process is determined.

[0030] During welding, if the local shrinkage vector in the predicted distribution map exceeds a preset threshold (e.g., 0.012m), the system automatically triggers an optimization process. It uses an iterative optimization algorithm to adjust the welding sequence parameters to obtain an optimized welding path plan. First, the system extracts local area data exceeding the threshold from the distribution map. For example, in an 80×80×25 grid within the welding area, the shrinkage value of a grid point near the middle of the weld reaches 0.0135m, significantly exceeding the threshold. The system marks this area as a high-risk deformation zone and records its coordinate range. Then, a genetic algorithm is used as the iterative optimization tool. The initial population size is set to 50, the crossover probability to 0.8, and the mutation probability to 0.05. The objective function is defined as minimizing the combined deviation between the local shrinkage value and the overall stress distribution. The maximum number of iterations is 100. The welding sequence parameters are adjusted through simulation calculations. For example, the welding sequence is changed from a linear path to a segmented alternating path, and the heat input is optimized from 2000W to 1600W, and the welding speed is adjusted from 0.004m / s to 0.0025m / s. Next, after each iteration, the system calls the finite element analysis module to recalculate the shrinkage distribution under the adjusted parameters and analyze whether the local shrinkage value has dropped below the threshold. For example, after the 50th iteration, the shrinkage value in the high-risk area dropped to 0.0118m, while the overall peak stress decreased by about 15%, meeting the optimization conditions. Finally, the system generates new welding path planning data based on the optimized parameters, refining the path point coordinates to 0.1mm accuracy, and automatically updates it to the welding control system to ensure that subsequent process execution conforms to the optimization results. The entire process is automated through Python scripts and the ANSYS interface, forming a closed-loop logic for calculation and path updates. To further improve the optimization effect, the system also stores the optimized data from each iteration in a database, forming a historical case library, so that reference parameters can be quickly called in subsequent similar scenarios, enhancing the system's adaptability.

[0031] S104. Extract the segment boundary coordinates from the optimized welding path plan, perform three-dimensional scanning data fusion on the curved surface structure, and determine the cumulative effect of dimensional deviation.

[0032] Segment boundary coordinate information is extracted from welding path planning data. Preliminary data processing is performed based on the geometric characteristics of the curved surface structure to obtain the basic dataset of segment boundaries. Using this basic dataset, combined with point cloud data of the curved surface structure obtained from 3D scanning, spatial alignment is performed using data fusion technology to determine the fused 3D structural data. Based on the fused 3D structural data, the geometric changes of the curved surface structure at different segment boundaries are analyzed to obtain the distribution data of dimensional deviations. For the dimensional deviation distribution data, if the deviation value exceeds a preset threshold at certain segment boundaries, the welding path planning data for that area is locally marked, resulting in a set of marked deviation regions. Using the set of marked deviation regions, combined with the overall geometric characteristics of the curved surface structure, a support vector machine algorithm is used to classify the deviation trends and determine the directional characteristics of deviation accumulation.

[0033]

[0034] Indicates the deviation trend classification result, This represents the trend feature vector of the deviation distribution data. Represents the kernel function. Represents support vectors, Represents the Lagrange multipliers. Indicates category label, This represents the number of support vectors. This represents the bias; this formula is used in the support vector machine algorithm to classify bias trends and determine the directional characteristics of bias accumulation.

[0035] Based on the directional characteristics of accumulated deviations, a local adjustment strategy for welding path optimization is generated, and the adjusted path planning data is determined. This adjusted path planning data is then remapped onto the 3D scanning data of the curved structure to obtain the final welding path and structure matching result.

[0036] After processing the optimized welding path planning, the system first extracts the segment boundary coordinates from the planning data. Assuming the total welding path length is 2.5m, divided into 5 segments, each 0.5m long, the system automatically identifies the start and end coordinates of each segment using a path parsing algorithm. For example, the first segment's start point is (0, 0, 0)mm, and the end point is (500, 10, 5)mm, accurate to 0.1mm. These coordinates are stored as a structured array for subsequent retrieval. Next, considering the characteristics of the curved surface structure, the system fuses the extracted boundary coordinates with 3D scanning data. The scanning data is acquired using a laser scanner, with a point cloud density set to 100 points per square centimeter, covering a curved surface area of ​​approximately 0.8m². The fusion process uses a point cloud registration algorithm (ICP algorithm), setting the maximum number of iterations to 30 and the convergence error threshold to 0.05mm. The spatial mapping relationship between the boundary coordinates and the scanning data is calculated. For example, if the actual position deviation of a certain boundary point is found to be 0.08mm, exceeding the expected tolerance of 0.03mm, the system automatically records the deviation data and generates a deviation distribution map. Finally, the system analyzes the cumulative effect of dimensional deviations using a Monte Carlo simulation method, with 1000 simulations. Assuming the deviation of each boundary segment follows a normal distribution with a standard deviation of 0.02 mm, the probability distribution of the cumulative deviation is calculated, revealing a 12% probability that the overall dimensional deviation could reach 0.15 mm. Combined with the geometric constraints of the curved surface structure, the system automatically generates deviation compensation coefficients, such as fine-tuning the boundary coordinates of the third path segment by -0.04 mm, forming a closed-loop data processing logic. To ensure the comprehensiveness of the analysis, the system also links the deviation data to the material property database of the curved surface structure, automatically extracting the material's elastic modulus (e.g., 210 GPa) as a reference input for subsequent deviation prediction, enhancing the accuracy of the analysis.

[0037] S105. Obtain the cumulative effect of the dimensional deviation after judgment, use the mesh deformation algorithm to simulate the assembly and fitting process, and determine the deviation correction matrix of the potential misalignment area.

[0038] The cumulative effect of dimensional deviations is assessed, and preliminary data extraction is performed on potential misalignment areas to obtain an initial distribution dataset of these areas. Based on this initial dataset, a mesh deformation algorithm is used to simulate the assembly process, determining the geometric deformation data of the misalignment areas during the simulation.

[0039]

[0040] This indicates the position of the geometric deformation of the k-th grid point during the simulated assembly and joining process. This represents the initial position of the k-th grid point. This represents the mesh deformation field function.

[0041] By analyzing geometric deformation data, potential misalignment areas are locally meshed, resulting in a set of mesh elements. For each mesh element, the deviation correction requirements during the assembly and mating process are analyzed to determine element-level deviation adjustment parameters. Based on these parameters, a deviation correction matrix for potential misalignment areas is constructed, yielding matrix-based correction data. This matrix-based correction data is then used to map the geometric deformation data during the assembly and mating process, determining the distribution of the corrected areas. If some mesh elements still exceed a preset threshold in the corrected distribution, secondary data extraction is performed on these elements to obtain the final deviation optimization dataset.

[0042] When dealing with the cumulative effect of dimensional deviations in curved structures, the system first retrieves the identified deviation data from the database. Assuming the overall structure contains 10 key connection points, with each point having a deviation value between 0.01mm and 0.12mm, the system uses a deviation accumulation analysis algorithm to integrate this data into a deviation vector matrix with dimensions of 10×3, corresponding to the deviation values ​​in the x, y, and z directions. For example, the deviation value of a certain point is (0.03, 0.05, 0.02)mm, accurate to 0.01mm. Subsequently, the system automatically calculates the sum of the cumulative deviations and finds that the maximum cumulative deviation is 0.09mm. The data is stored as a high-precision floating-point array for subsequent retrieval. Next, the system uses a mesh deformation algorithm to simulate the assembly process. Specifically, the curved surface structure is divided into 1000 mesh elements, each with an area of ​​approximately 0.01 m². Using the finite element analysis (FEA) algorithm, the mesh deformation constraint is set to a maximum deformation of 0.1 mm and 50 iterations. The system calculates the stress distribution of each mesh element during the assembly process. For example, if the stress peak in a certain area is 15 MPa, exceeding the safety threshold of 10 MPa, the system automatically marks this area as a potential misalignment area and generates a stress distribution cloud map. Finally, the system determines the deviation correction matrix for potential misalignment areas and uses an inverse deformation calculation method, setting the matrix dimension to 3×3, to calculate the correction coefficient for each misalignment area. For example, the correction coefficient for a certain area is (1.02, 0.98, 1.01), accurate to 0.01. The system automatically adjusts the position of the mesh cells based on the geometric boundary conditions of the curved surface structure. For example, the center point coordinates of a certain cell are adjusted from (100.5, 200.3, 50.2) mm to (100.6, 200.2, 50.3) mm. At the same time, the correction matrix is ​​associated with the material thermal expansion coefficient database, and the thermal expansion coefficient of 11.5×10^-6 / ℃ is extracted as an auxiliary parameter of the correction matrix to ensure the rigor of the deviation correction logic and form a complete data processing closed loop.

[0043] S106. By applying a virtual force field constraint to the segmented model using the determined deviation correction matrix, the corrected segmented geometry is obtained.

[0044] Using the generated deviation correction matrix, virtual force field constraints are applied to the geometric morphology data of the segmented model to obtain a preliminary corrected segmented geometry. Based on the preliminary corrected segmented geometry, morphological adjustment data for each segment element is extracted to determine the local deformation of each element under force field constraints. For the local deformation, the geometric morphological changes of each segment element under constraints are analyzed to obtain the changed element morphological distribution data. Using the changed element morphological distribution data, a morphological adjustment mapping for the segmented model is constructed, and the consistency of the overall geometric morphology after mapping is judged. If the overall geometric morphological consistency does not reach a preset threshold, a secondary allocation of local force field constraints is performed on the substandard segment elements to obtain adjusted element morphological data. Based on the adjusted element morphological data, the corrected morphological information of the segmented model is updated to determine the final geometric morphological correction result. Using the final geometric morphological correction result, the data of each element in the segmented model is integrated to generate a complete corrected morphological dataset.

[0045] Specifically, when processing the geometric correction of the segmented surface model, the system applies virtual force field constraints to the geometric data of the segmented model based on the generated deviation correction matrix.

[0046] For example, for a large curved surface segment containing 500 nodes, the system applies a virtual tensile and compressive stress field with a magnitude between 50 and 150 Newtons in its boundary connection region. By simulating the elastic mechanical response of the material, the nodes are driven to undergo small displacements, thereby obtaining a preliminary corrected segment geometry.

[0047] It should be noted that, based on the preliminary corrected segmented geometry, the system will extract the morphological adjustment data for each segment unit.

[0048] For example, the system records that a key mesh element under the action of a virtual force field has a displacement of 0.05 mm along the normal direction of the surface and a tangential displacement of 0.02 mm. This allows the system to determine the local deformation of each element under the force field constraint, including the microscopic changes in principal curvature and Gaussian curvature.

[0049] Specifically, for cases of local deformation, the system further analyzes the geometric changes of each segmented unit under constraints.

[0050] For example, by comparing the node coordinate matrices before and after stress, the system quantitatively analyzes the change in surface distortion and obtains the changed unit morphology distribution data. This data is stored in a structured form as a set of three-dimensional spatial coordinates. Using the changed unit morphology distribution data, the system constructs a morphology adjustment mapping for the segmented model, transforming the deformation in the local coordinate system to the global assembly coordinate system, and then judging the consistency of the overall geometric shape after mapping.

[0051] For example, the system calculates the angle between the normal vectors of corresponding nodes at the boundary of adjacent segments. If it finds that the angle reaches 0.8 degrees, which exceeds the preset threshold of 0.5 degrees, it determines that the overall geometric consistency has not reached the preset threshold.

[0052] It should be noted that if the consistency standard is not met, the system will perform a secondary allocation of local force field constraints on the segmented units that do not meet the standard.

[0053] For example, for regions where the included angle of the normal vectors exceeds the limit, the system superimposes a 20-Newton compensation force field on top of the original virtual force field and adjusts the direction of the force field. After recalculation, the adjusted element morphology data is obtained. Based on the adjusted element morphology data, the system updates the corrected morphology information of the segmented model, overwriting the original data with the corrected node coordinates, and determining the final geometric morphology correction result. Using the final geometric morphology correction result, the system integrates the data of each element in the segmented model, merging the spatial coordinates and curvature parameters of all nodes to generate a complete corrected morphology dataset.

[0054] S107. Based on the corrected segmented geometry, generate a digital twin model for assembly alignment, determine whether the overall accuracy meets the preset standard, and obtain the final manufacturing adjustment plan.

[0055] By correcting the shape data, the boundary point set of the segmented geometry is extracted, and spatial position mapping is performed on the boundary point set to obtain the initial frame data for assembly alignment.

[0056]

[0057] This represents the initial frame data for assembly alignment. Represents the rotation matrix. Represents the translation vector. Indicates the number of boundary point sets. Represents the piecewise geometric boundary. One point, Represents a reference assembly point, This represents the Euclidean norm; this formula maps the spatial positions of the boundary point set and obtains the initial frame data by minimizing the mean square distance between the transformed points and the reference point.

[0058] Based on the initial framework data, a virtual assembly environment of the digital twin is constructed, and the corrected shape information of the segmented geometry is loaded to determine the preliminary distribution of overall accuracy. For this preliminary distribution, specific areas of deviation distribution are analyzed, and a preset threshold is used for comparison to determine whether the overall accuracy meets the standard range. If the overall accuracy does not reach the preset threshold, a set of parameters for local adjustments is generated for areas with large deviations, and the priority order of manufacturing adjustments is determined. Based on the priority order, adjustment instructions for the preliminary plan are generated, and the digital twin model is locally updated in virtual assembly based on the adjustment criteria to obtain the adjusted accuracy distribution data. Using the adjusted accuracy distribution data, the improvement in overall accuracy is analyzed, and a support vector machine algorithm is used to classify the deviation distribution to determine the final manufacturing adjustment plan.

[0059]

[0060] This represents the classification decision function of the support vector machine. This represents the deviation feature vector of the adjusted precision distribution data. Indicates the number of support vectors. Indicates the first Lagrange multipliers of the support vectors, Indicates the first The class labels of the support vectors. Represents the kernel function. This represents the bias term; this formula uses a support vector machine algorithm to classify the bias distribution and determine the final manufacturing adjustment plan.

[0061] Based on the final manufacturing adjustment plan, assembly alignment guidance data for segmented geometry is generated, the virtual environment information of the digital twin model is updated, and a complete adjusted dataset is obtained.

[0062] The entire process of generating a digital twin model for assembly alignment based on the corrected segmented geometry, and determining whether the overall accuracy meets preset standards, can be automated through information technology to obtain the final manufacturing adjustment plan. First, for the corrected segmented geometry data, the system automatically imports the geometric data of each segment using 3D modeling software. For example, a segment with a length of 100.25 mm, a width of 50.15 mm, and a height of 30.10 mm is used. The relative positional deviation between each segment is calculated using a point cloud matching algorithm. Assuming a deviation value of 0.05 mm, the system compares this deviation with a preset tolerance of 0.1 mm to confirm whether adjustment is needed. Next, when generating the digital twin model, the system uses finite element analysis to simulate the assembly process, calculating the stress distribution of each segment in the virtual environment. For example, the maximum stress value is 25.5 MPa, lower than the material's allowable value of 30 MPa, indicating structural stability. Then, when judging the overall accuracy, the system uses a global error analysis algorithm to sum the assembly errors of all segments, obtaining a total error of 0.08 mm, which is lower than the preset standard of 0.1 mm, thus determining that the accuracy is acceptable. If the error exceeds the standard, for example, reaching 0.12 mm, the system will automatically trigger the adjustment plan generation module. Using optimization algorithms (such as genetic algorithms), it calculates the optimal adjustment parameters. For example, adjusting the offset of a certain segment by 0.03 mm, after re-simulating assembly, the error drops to 0.09 mm, meeting the standard. Finally, the system generates a detailed numerical report of the manufacturing adjustment plan, for example, adjusting the machining allowance of a certain segment to 0.02 mm, and outputs the plan to the manufacturing execution system, interfacing with the production equipment to achieve automated adjustment. Through this process, the data at each stage are closely linked, forming a complete logical chain from geometric correction to accuracy judgment to adjustment plan, ensuring the efficiency and accuracy of the technical implementation.

[0063] The above description is merely a preferred embodiment of one or more embodiments of this specification and is not intended to limit the scope of one or more embodiments of this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of one or more embodiments of this specification should be included within the protection scope of one or more embodiments of this specification.

Claims

1. A method for compensating for welding shrinkage and assembly deviations in the segmented manufacturing of marine suction sails, characterized in that, The method includes: A thermodynamic segmented model of the welding process is constructed using the finite difference method. The temperature distribution field is obtained from the preset material properties and welding parameters, and the initial shrinkage vector of each segment is obtained. Based on the obtained initial shrinkage vector, a prediction model is trained using the support vector machine algorithm, and features are extracted from historical welding data to determine the predicted distribution map of shrinkage deformation. If the local shrinkage vector in the predicted distribution map exceeds the preset threshold, the welding sequence parameters are adjusted through an iterative optimization algorithm to obtain an optimized welding path plan. Segment boundary coordinates are extracted from the optimized welding path plan, and three-dimensional scanning data is fused for the curved structure to determine the cumulative effect of dimensional deviations. The cumulative effect of dimensional deviations after judgment is obtained, and the assembly and fitting process is simulated using a mesh deformation algorithm to determine the deviation correction matrix for potential misalignment areas. By applying a virtual force field constraint to the piecewise model using a determined deviation correction matrix, the corrected piecewise geometry is obtained. Based on the corrected segmented geometry, a digital twin model for assembly alignment is generated to determine whether the overall accuracy meets the preset standards, and the final manufacturing adjustment plan is obtained.

2. The method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to claim 1, characterized in that, The thermodynamic segmented model of the welding process is constructed using the finite difference method. The temperature distribution field is obtained from preset material properties and welding parameters to obtain the initial shrinkage vector for each segment, including: By using preset material properties and welding parameters, the welding process is meshed using the finite difference method, a thermodynamic segmented model is constructed, and the heat conduction equations of each mesh element are obtained. Based on the heat conduction equations, the temperature values ​​of each grid element at different time steps are calculated iteratively to obtain the dynamic temperature distribution field during the welding process. ； Represents grid points , The temperature value at time step n. Indicates the thermal diffusivity. Indicates the time step. This represents the spatial grid spacing, and the formula is used to iteratively calculate the temperature distribution field of each grid cell at different time steps. For the dynamic temperature distribution field, the temperature gradient changes in each segment are analyzed to determine the correspondence between temperature gradient and thermal stress. By calculating the thermal stress distribution in each segment through the correspondence between temperature gradient and thermal stress, preliminary vector data of initial shrinkage are derived. The finite difference method is used to correct the boundary conditions of the preliminary vector data to obtain the accurate initial shrinkage vector of each segment. ； Represents the exact initial contraction vector. This represents the initial contraction vector. Representing the normal derivative, This formula represents the boundary length and is used in the finite difference method to correct the boundary conditions of the initial vector data, thereby obtaining the accurate initial shrinkage vector for each segment. If the calculated result of the precise initial shrinkage vector exceeds the preset threshold range, the mesh density is adjusted, the temperature distribution field is recalculated, and the rationality of the final shrinkage vector is judged. By using the adjusted temperature distribution field and contraction vector, thermodynamic response data for each segment are generated to determine the overall deformation trend of the welding process.

3. The method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to claim 1, characterized in that, The step of training a prediction model using a support vector machine algorithm based on the obtained initial shrinkage vector, extracting features from historical welding data, and determining the predicted distribution map of shrinkage deformation includes: By using initial shrinkage and vector data, a prediction model based on the support vector machine algorithm is constructed. Historical data is then processed to obtain a structured dataset after feature extraction. Based on the structured dataset after feature extraction, the support vector machine algorithm is used to train the model and determine the preliminary predicted distribution of shrinkage deformation. ； This indicates a preliminary predicted distribution. Indicates the number of support vectors. Represents the Lagrange multipliers. Indicates historical data tags, Represents the kernel function. This represents a structured dataset sample after feature extraction. Indicates input features, This represents the bias term; this formula is used to determine the preliminary predicted distribution of shrinkage deformation based on the structured dataset after feature extraction using the support vector machine algorithm for model training; Based on the preliminary predicted distribution, data from key time points in the welding process are obtained to determine the phased changes in the deformation trend; If the fluctuation of the deformation trend at a certain time point exceeds the preset threshold, the historical data is segmented to obtain the adjusted feature dataset. By reloading the adjusted feature dataset into the prediction model, the updated shrinkage deformation distribution map is determined. Obtain the updated shrinkage deformation distribution map and, in conjunction with the process parameters of the welding process, determine the stability of the overall deformation trend; Based on the stability of the overall deformation trend, staged deformation prediction data for the welding process are generated to determine the final distribution prediction results.

4. The method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to claim 1, characterized in that, If the local shrinkage vector in the predicted distribution map exceeds a preset threshold, the welding sequence parameters are adjusted through an iterative optimization algorithm to obtain an optimized welding path plan, including: By using the data from the predicted distribution map, we can obtain the specific distribution of local contraction vectors and determine the range of areas exceeding the preset threshold. For regions exceeding a preset threshold, an iterative optimization algorithm is used to process the parameter configuration of the welding sequence, resulting in an adjusted parameter set. ； This represents the adjusted set of parameters. This indicates the welding sequence parameter configuration. This indicates the range of regions exceeding the threshold. Represents a parameter-based local contraction vector. Indicates points within the region; Based on the adjusted parameter set, a preliminary optimized path scheme is generated, and the applicability of the path scheme in distribution analysis is determined. If there are local mismatches in the path scheme during distribution analysis, the welding plan is segmented to obtain the segmented path data. The segmented path data is reloaded into the iterative optimization algorithm to determine the updated welding sequence scheme; The updated welding sequence scheme is obtained, and combined with the results of the distribution analysis, the final welding planning scheme is generated. For the final welding plan, output complete optimized path data and determine the sequential execution logic in the welding process.

5. The method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to claim 1, characterized in that, The process of extracting segmented boundary coordinates from the optimized welding path plan, performing 3D scanning data fusion on the curved surface structure, and determining the cumulative effect of dimensional deviations includes: The segment boundary coordinate information is extracted from the welding path planning data, and preliminary data processing is carried out based on the geometric characteristics of the curved surface structure to obtain the basic dataset of the segment boundary. Using the basic dataset of segmented boundaries, combined with the surface structure point cloud data obtained by 3D scanning, spatial alignment processing is performed using data fusion technology to determine the fused 3D structure data; Based on the fused 3D structural data, the geometric changes of the curved surface structure at different segment boundaries are analyzed to obtain the distribution data of dimensional deviations; For the distribution data of dimensional deviations, if the deviation value exceeds the preset threshold at certain segment boundaries, the welding path planning data of that area is locally marked to obtain a set of marked deviation areas. By using the marked set of deviation regions and combining the overall geometric characteristics of the surface structure, the support vector machine algorithm is used to classify the deviation trend and determine the directional characteristics of the deviation accumulation. ； Indicates the deviation trend classification result, This represents the trend feature vector of the deviation distribution data. Represents the kernel function. Represents support vectors, Represents the Lagrange multipliers. Indicates category label, This represents the number of support vectors. This indicates the bias; this formula is used in the support vector machine algorithm to classify bias trends and determine the directional characteristics of bias accumulation. Based on the directional characteristics of accumulated deviations, a local adjustment strategy for welding path optimization is generated, and the adjusted path planning data is determined. The adjusted path planning data is remapped into the 3D scanning data of the curved structure to obtain the final welding path and structure matching result.

6. The method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to claim 1, characterized in that, The cumulative effect of the obtained and judged dimensional deviations is simulated using a mesh deformation algorithm to determine the deviation correction matrix for potential misalignment regions, including: Obtain the cumulative effect judgment results of size deviation, perform preliminary data extraction for potential misalignment areas, and obtain the initial distribution dataset of misalignment areas; Based on the initial distribution dataset, a mesh deformation algorithm is used to simulate the assembly and mating process to determine the geometric deformation data of the misaligned area during the simulation process. ； This indicates the position of the geometric deformation of the k-th grid point during the simulated assembly and joining process. This represents the initial position of the k-th grid point. Represents the mesh deformation field function; Using geometric deformation data, local meshing is performed on potential misalignment areas to obtain a set of mesh cells. For a set of mesh elements, analyze the deviation correction requirements of each element in the assembly and mating process, and determine the deviation adjustment parameters at the element level; Based on the deviation adjustment parameters at the unit level, a deviation correction matrix for potential misalignment areas is constructed to obtain matrix-based correction data. By using matrix-based correction data, the geometric deformation data in the assembly and fitting process is mapped and processed to determine the distribution of the corrected areas. If some grid cells in the corrected regional distribution still exceed the preset threshold, then secondary data extraction is performed on these cells to obtain the final deviation optimization dataset.

7. The method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to claim 1, characterized in that, The process of applying virtual force field constraints to the piecewise model using a determined deviation correction matrix to obtain the corrected piecewise geometry includes: By applying virtual force field constraints to the geometric data of the segmented model using the generated deviation correction matrix, a preliminary corrected segmented geometry is obtained. Based on the preliminarily corrected segmented geometry, extract the shape adjustment data of each segmented unit and determine the local deformation of each unit under force field constraints. For local deformation, the geometric shape changes of each segmented element under constraints are analyzed, and the distribution data of the changed element shape are obtained. By using the changed unit morphology distribution data, a morphology adjustment mapping for the segmented model is constructed, and the consistency of the overall geometric morphology after mapping is judged. If the overall geometric consistency does not reach the preset threshold, the substandard segmented units are subjected to a secondary allocation of local force field constraints to obtain the adjusted unit shape data. Based on the adjusted unit morphology data, update the correction morphology information of the segmented model and determine the final geometric morphology correction result. Based on the final geometric morphology correction results, the data of each unit of the segmented model are integrated to generate a complete corrected morphology dataset.

8. The method for compensating for welding shrinkage and assembly deviation in the segmented manufacturing of marine suction sails according to claim 1, characterized in that, The process of generating a digital twin model for assembly alignment based on the corrected segmented geometry, determining whether the overall accuracy meets preset standards, and obtaining the final manufacturing adjustment plan includes: By correcting the shape data, the boundary point set of the segmented geometry is extracted, and the spatial position is mapped for the boundary point set to obtain the initial frame data for assembly alignment. ； This represents the initial frame data for assembly alignment. Represents the rotation matrix. Represents the translation vector. Indicates the number of boundary point sets. Represents the piecewise geometric boundary. One point, Represents a reference assembly point, denoted by Euclidean norm; this formula performs spatial location mapping on the boundary point set, and obtains the initial frame data by minimizing the mean square distance between the transformed points and the reference points; Based on the initial framework data, a virtual assembly environment of digital twin is constructed, the correction shape information of segmented geometry is loaded, and the preliminary distribution state of overall accuracy is determined. Based on the initial distribution status, the specific areas of the deviation distribution are analyzed, and a preset threshold is used for comparison to determine whether the overall accuracy meets the standard range. If the overall accuracy does not reach the preset threshold, a set of parameters for local adjustment is generated for areas with large deviation distribution to determine the priority order of manufacturing adjustments. Based on the priority order, adjustment instructions for the preliminary plan are generated. Combined with the adjustment criteria, the digital twin model is partially updated by virtual assembly to obtain the adjusted accuracy distribution data. By analyzing the overall improvement in accuracy based on the adjusted accuracy distribution data, the deviation distribution is classified using the support vector machine algorithm to determine the final manufacturing adjustment plan. ； This represents the classification decision function of the support vector machine. This represents the deviation feature vector of the adjusted precision distribution data. Indicates the number of support vectors. Indicates the first Lagrange multipliers of the support vectors, Indicates the first The class labels of the support vectors. Represents the kernel function. This represents the bias term; this formula uses a support vector machine algorithm to classify the bias distribution and determine the final manufacturing adjustment plan. Based on the final manufacturing adjustment plan, assembly alignment guidance data for segmented geometry is generated, the virtual environment information of the digital twin model is updated, and a complete adjusted dataset is obtained.

Images