Welding multi-scale simulation and optimization method for nuclear fusion device

Abstract

The application discloses a welding multi-scale simulation and optimization method of a nuclear fusion device, and aims at solving the problems of the application of the existing welding simulation technology in large nuclear fusion devices and matched assembly tool structures, and provides an efficient simulation method based on a welding seam segmented activation strategy, multi-scale finite element analysis technology and Abaqus+Fortran+Python collaborative control, which comprises multi-scale modeling and domain division, welding heat source modeling and heat input definition, transient heat conduction and material thermal property modeling, thermal-mechanical coupling analysis and stress field solving, welding seam segment model automatic generation and heat history transmission, stress / deformation result splicing and full structure response reconstruction, and Python automatic control process and expansion interface. The method can significantly reduce the calculation resource consumption and simulation time of the overall model under the premise of ensuring the local response prediction accuracy of the welding seam, and has good expansibility and engineering deployment capacity.

Description

Technical Field

[0001] This invention relates to the fields of welding technology and computer-aided engineering simulation technology, specifically to a multi-scale simulation and optimization method for welding of nuclear fusion devices. Background Technology

[0002] The development and construction of large-scale nuclear fusion devices involve numerous welding operations on complex structures, such as vacuum chambers, support structures, cooling pipelines, magnet support components, and large assembly tools. These components are typically large in size, complex in structure, and require extremely high positioning accuracy. The welding quality of these components directly affects the structural integrity, operational stability, and safety performance of the device.

[0003] The heat input generated during welding can cause localized non-uniform temperature rises, leading to thermal expansion and plastic deformation, resulting in welding deformation and residual stress in the workpiece. These welding effects not only affect the accuracy of subsequent assembly but can also, in severe cases, induce engineering problems such as structural cracking and stress corrosion damage. Therefore, scientifically predicting deformation trends and stress distribution before welding and optimizing welding process parameters accordingly is crucial to ensuring the manufacturing quality and operational reliability of the equipment.

[0004] Currently, two main methods are used in engineering to evaluate welding results and formulate processes:

[0005] (1) Experimental method: Welding tests are conducted in the laboratory under laboratory conditions by making scaled-down or equal-scale test pieces to measure post-weld deformation and residual stress. Although this method is reliable, it is time-consuming and costly, and the test samples are difficult to fully represent the complex working conditions and boundary conditions in the actual structure.

[0006] (2) Experience and Standards Method: Welding parameters and process routes are formulated based on past welding engineering experience and industry standards. This method relies on experience accumulation, lacks scientific and visual predictive capabilities for specific structures, and is difficult to adapt to the structural design requirements of high complexity and high precision in nuclear fusion devices.

[0007] In terms of computational simulation, although existing CAE tools (such as Abaqus) have the function of welding thermal-structure coupling simulation, which can simulate the heat conduction, thermal expansion and stress development process during welding, when the structure size is large or the weld seam is densely distributed, traditional simulation methods need to divide the welding area into high-density meshes to ensure calculation accuracy, which leads to a surge in the degree of freedom of the model, resulting in huge consumption of computing resources and extremely long solution time, making it difficult to widely deploy in engineering practice.

[0008] Therefore, there is an urgent need for a method that can balance simulation accuracy and computational efficiency, be applicable to the prediction of welding deformation and residual stress in large and complex structures, and support multi-parameter welding process optimization and visualization feedback, so as to improve the welding quality control level and manufacturing intelligence level of nuclear fusion devices.

[0009] Currently, the following technical solutions are mainly adopted in engineering practice and research for modeling and optimizing the welding process of complex structures in large-scale nuclear fusion devices:

[0010] (1) Single-scale high-precision finite element welding simulation method

[0011] This method, based on general-purpose finite element software (such as Abaqus and ANSYS), uses explicit or implicit thermo-mechanical coupling analysis to model the welding process step by step, apply heat input loads, and simulate the phase transformation behavior of metals, thereby obtaining the temperature field, deformation field, and residual stress field of the structure after welding. To improve simulation accuracy, fine meshes are typically generated in the weld region.

[0012] defect:

[0013] a) When the structural size is large (such as the main structure of a nuclear fusion device), in order to ensure the simulation accuracy of the welding area, the number of fine meshes will increase significantly, resulting in an extremely high number of degrees of freedom, and the time and resource costs required for computation will increase exponentially.

[0014] b) It lacks multi-scale control capabilities, making it difficult to simplify the modeling of non-critical areas while ensuring high accuracy in critical areas;

[0015] c) It is difficult to apply to scenarios involving multi-scheme comparison, parameter sensitivity analysis, and engineering iterative optimization.

[0016] (2) Scaled-down experiment and experimental fitting prediction method

[0017] This method involves preparing scaled-down test specimens to reproduce actual welding conditions. Three-dimensional deformation measurement technology and stress-strain testing methods (such as X-ray diffraction, hole drilling, DIC measurement, etc.) are used to determine the deformation and residual stress after welding. The test results are then numerically fitted to guide the process design of full-size workpieces.

[0018] defect:

[0019] a) Due to scaling errors and boundary effects, the response of the test specimen cannot fully represent the actual structural behavior;

[0020] b) Each iteration of the design requires the creation of new test samples, resulting in long experimental cycles and high costs;

[0021] c) The results have low visualization capabilities and cannot achieve real-time prediction of welding behavior across the entire structure.

[0022] (3) Process formulation method based on empirical parameters

[0023] In industrial manufacturing, especially in nuclear equipment manufacturing enterprises, the process flow, such as welding sequence, welding path, and fixture fixing method, is often formulated primarily based on experience and secondarily on standards. For example, the heat input range and welding speed are selected according to standards such as GB / T 985 and ASME BPVC, as well as experience from past projects.

[0024] defect:

[0025] a) Lacks the ability to scientifically assess structural response;

[0026] b) Difficult to adapt to complex weld layouts and extreme precision control requirements;

[0027] c) Design requirements that are not suitable for new structures, new materials, or extreme service environments. Summary of the Invention

[0028] The present invention proposes a method, equipment and storage medium for multi-scale simulation and optimization of welding of nuclear fusion devices, which can at least solve one of the technical problems in the background art.

[0029] To achieve the above objectives, the present invention adopts the following technical solution:

[0030] A multi-scale simulation and optimization method for welding in nuclear fusion devices involves executing the following steps using computer equipment.

[0031] Step 1: Multi-scale modeling and domain partitioning

[0032] First, based on the actual geometric model of the nuclear fusion device or tooling, the overall structure is divided into regions. The core of this step is to establish a multi-scale finite element model: welds and heat-affected zones are modeled using fine meshes to ensure local calculation accuracy; the main structure far from the weld area is modeled using coarse meshes to reduce the overall number of degrees of freedom. Continuous coupling between coarse and fine meshes is achieved using the MPC or Tie method in Abaqus, thus obtaining a multi-scale model that balances computational efficiency and accuracy. The output of this step is a basic finite element analysis model containing global and local scales, providing geometric and mesh support for subsequent heat source loading and physics field calculations.

[0033] Step 2: Welding heat source modeling and heat input definition

[0034] After completing the geometry and mesh generation, the next step is to physically model the energy input of the welding process. The Goldak double-ellipsoidal heat source model is used to describe the heat input distribution, and its dynamic movement and loading along the weld path is controlled by the user subroutine DFLUX written in Fortran. The core technology of this step is to achieve a realistic physical representation of the welding heat source. Its input is the multi-scale geometric model and weld path data from step 1, and its output is the heat input distribution field that varies with time and space, laying the physical driving conditions for subsequent temperature field calculations.

[0035] Step 3: Transient heat conduction and material thermal property modeling

[0036] Based on the heat source model, a transient heat conduction equation for the welding process is established, and temperature-varying material thermophysical parameters, such as density, specific heat capacity, thermal conductivity, and latent heat of phase change, are introduced. The core of this step is to obtain the time-varying temperature field distribution through numerical solution. Its input is the heat input distribution defined in step 2, and the output is the evolution result of the temperature field, while providing boundary conditions and driving forces for subsequent thermo-mechanical coupling analysis.

[0037] Step 4: Thermo-mechanical coupling analysis and stress field solution

[0038] After obtaining the temperature field distribution, a thermo-mechanical coupling calculation is further performed. The core of this step is based on the thermo-elastic-plastic constitutive relationship, considering thermal expansion, material plasticity, and yielding behavior, to obtain the stress and deformation fields that evolve over time. The input is the temperature field from step 3, and the output is the residual stress distribution and transient deformation results, providing a mechanical response basis for the continuous calculation of the weld segment model and subsequent splicing.

[0039] Step 5: Automatic generation of weld segment model and heat history transfer

[0040] For complex long welds, a Python script is used to segment the weld path, generating an independent sub-model for each weld segment and automatically submitting it for calculation. The core technology of this step is the thermal history transfer mechanism: the boundary node temperature obtained from the previous welding simulation is used as the initial thermal boundary condition for the next segment, achieving continuity of the welding thermal history. The input is the temperature and stress results calculated in step 4, and the output is the welding effect data of each sub-model segment, providing distributed simulation units for global stitching.

[0041] Step 6: Stress / deformation result splicing and full structural response reconstruction

[0042] After completing the independent calculations for each weld segment, the results of the sub-models are stitched together into the global structure using methods such as stress projection, mesh interpolation, or thermal field reconstruction. The core of this step is establishing a mapping relationship between the segmented simulations and the overall structural response, ensuring the physical continuity and spatial consistency of the global residual stress field and the welding deformation field. The input is the simulation results of each segment from step 5, and the output is a visualized reconstruction of the global residual stress distribution and overall welding deformation, providing ultimately usable data support for engineering applications.

[0043] Step 7: Python Automatic Control Flow and Extended Interface

[0044] Finally, Python scripts are used to automate the entire process of scheduling and data management, including sub-model construction, heat source path loading, simulation task submission, result extraction, and inter-segment data transmission. The core technology of this step lies in process integration and open interfaces, which not only significantly reduces manual operation and debugging time but also provides extended interfaces for subsequent introduction of optimization algorithms (such as genetic algorithms and Bayesian optimization), process database construction, and intelligent process recommendation. The input is the analysis process of the first six steps, and the output is standardized simulation results and visualized data, ultimately forming a reusable and scalable multi-scale welding simulation and optimization framework.

[0045] As can be seen from the above technical solution, the multi-scale simulation and optimization method for welding of nuclear fusion devices of the present invention relates to the fields of welding process and computer-aided engineering (CAE) simulation technology. Specifically, it is a method for predicting and optimizing welding deformation and residual stress in large-scale nuclear fusion devices and their related assembly tooling structures. This method integrates multi-scale finite element modeling, thermo-mechanical coupling analysis, Python program control, and data processing. It is suitable for the formulation of welding processes and optimization of welding parameters for large and complex structures, and belongs to the interdisciplinary field of nuclear fusion engineering structure manufacturing and high-end manufacturing simulation technology.

[0046] This invention aims to solve the following key technical problems faced in the welding process design and quality control of current large-scale nuclear fusion devices and their assembly tooling structures:

[0047] ① The contradiction between welding simulation accuracy and computational resources

[0048] In traditional welding finite element simulation, to obtain accurate predictions of welding deformation and residual stress, it is necessary to divide the weld region into fine meshes and apply heat input simulation for each weld pass. However, when dealing with large nuclear fusion structures (such as vacuum chambers or support fixtures with diameters of several meters and weld lengths of tens of meters), using a uniform fine mesh for the entire structure would lead to a dramatic increase in the number of meshes, with degrees of freedom reaching millions or even tens of millions. The computation time would be unacceptable, and the simulation resource consumption would far exceed the tolerance of engineering applications, leading to the abandonment of such simulations in practical engineering.

[0049] ② Welding process evaluation methods lack systematicity and visualization capabilities.

[0050] Currently, most welding process development relies on accumulated experience or localized scaled-down experiments, lacking visual prediction and verification methods for the overall structural response. Engineers find it difficult to intuitively understand the impact of welding sequence, heat input strategies, or fixture arrangement on the overall structural response (such as deformation direction and stress concentration areas), limiting the efficiency of welding scheme optimization and the level of quality control.

[0051] ③ Lack of simulation platforms that can support automated parameter control and optimization.

[0052] Welding results are influenced by a combination of factors, such as welding path, heat source distribution, welding speed, and cooling method. Traditional simulation processes involving manual parameter setting and repetitive modeling are inefficient, making it difficult to conduct systematic evaluations and explore optimization strategies for multi-parameter combined processes. Furthermore, there is a lack of a unified and integrated simulation-analysis-optimization toolchain.

[0053] ④ Nuclear fusion devices have complex structures and diverse boundary conditions, making it difficult to adapt to general simulation processes.

[0054] Nuclear fusion devices contain a variety of large thin-walled structures, dissimilar material joints, and complex fixture systems. Existing simulation processes lack general adaptability for such multi-condition scenarios, and cannot support batch simulation evaluation and process reuse of diverse structures, which seriously restricts the digital transformation of welding and the progress of intelligent manufacturing.

[0055] This invention proposes a multi-scale simulation and optimization method for welding deformation and residual stress in large-scale nuclear fusion devices. It overcomes the problems of traditional welding simulation methods, such as huge computational costs, cumbersome modeling processes, invisibility of response results, and inability to perform parameter optimization, when dealing with large and complex structures. It has the following significant advantages:

[0056] (1) Balancing simulation accuracy and computational efficiency to achieve efficient welding simulation of large structures

[0057] Compared to traditional global fine mesh modeling, this invention significantly reduces the total degrees of freedom of the model by combining local fine modeling of the weld area with coarsening of non-critical areas through a multi-scale strategy. Furthermore, by using a segmented activation of thermal input along the weld path, the solution is performed only on local areas, thus avoiding the problem of computational resource explosion in transient coupled thermo-mechanical simulation of the entire structure.

[0058] (2) Supports physical simulation of weld heat input, improving the accuracy of prediction results.

[0059] By embedding a Fortran user subroutine (DFLUX) based on the Goldak double ellipsoid model into the Abaqus platform, dynamic movement and loading of the heat source along the weld path are achieved, accurately representing the non-uniform volumetric heat source distribution during the welding process. Compared to traditional surface heat sources or simplified temperature loading methods, the method of this invention can more realistically reproduce the welding thermal cycle, weld penetration characteristics, and heat-affected zone evolution process, thereby significantly improving the physical reliability of temperature and stress field simulations.

[0060] (3) Introduce automated modeling and inter-segment thermal history coupling to improve the flexibility and scalability of the simulation process.

[0061] This invention automates the entire process of weld segmentation, sub-model generation, thermal boundary extraction, and simulation scheduling through Python control scripts. This allows engineers to flexibly set simulation accuracy and calculation scale according to the actual complexity of the welded structure. At the same time, it automatically completes the seamless transfer of temperature fields between adjacent segments, ensuring the temporal continuity and spatial coupling of the welding thermal history, and significantly reducing the time and workload of manual modeling and parameter tuning.

[0062] (4) Supports the splicing and visualization integration of multi-segment weld response results to enhance engineering evaluation capabilities.

[0063] This invention designs a strategy for inter-segment splicing and full-structure mapping of stress / temperature results. Combined with a Python post-processing module, it can output spatial distribution maps of welding residual stress and welding shrinkage deformation in each segment, and support three-dimensional visualization of the results. This greatly enhances engineers' ability to understand and judge welding quality and structural response, and helps guide process decisions and structural design optimization.

[0064] (5) It has open interfaces to support subsequent process optimization and intelligent recommendation extension.

[0065] This invention pre-defines parameter control interfaces and structured result output formats in the simulation process, providing a data foundation for subsequent introduction of algorithm modules such as welding path optimization, heat input scheduling optimization, and fixture scheme optimization. Simultaneously, the accumulation of simulation result libraries enables the construction of a knowledge-driven welding process database, thereby serving future intelligent manufacturing and digital twin system architectures.

[0066] (6) It can be widely applied to various types of welded structures in nuclear fusion devices, and has strong engineering versatility.

[0067] The feasibility of this invention's method has been verified in several typical structures, including vacuum chamber circumferential seams, spiral magnet support flange welds, and cooling pipe support fillet welds, all demonstrating good adaptability and accuracy. Its modeling logic and control mechanism are highly modular and parameterized, facilitating rapid migration and reuse, and are suitable for unified simulation evaluation and process management of batch-welded structures. Attached Figure Description

[0068] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0069] Figure 2 This is a schematic diagram of the local residual stress distribution in the segmented welding area provided in an embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of the local welding deformation distribution in the segmented welding area provided in an embodiment of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0072] like Figure 1 As shown in this embodiment, the welding multi-scale simulation and optimization method for nuclear fusion devices addresses several limitations of existing welding simulation technologies in the application of large-scale nuclear fusion devices and their supporting assembly tooling structures. These limitations include high overall modeling resource consumption, the trade-off between simulation accuracy and computational efficiency, and the difficulty in accurately visualizing and predicting welding effects. This method proposes an efficient simulation approach based on a weld segment activation strategy, multi-scale finite element analysis technology, and collaborative control using Abaqus + Fortran + Python. This approach can significantly reduce the overall model's computational resource consumption and simulation time while ensuring the accuracy of local weld response prediction, and also possesses good scalability and engineering deployment capabilities.

[0073] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0074] The method of this invention follows a logical chain of "geometry—physics—numerical—data," gradually realizing a complete closed loop from structural modeling to global welding effect prediction. Specifically, it includes the following steps:

[0075] Step 1: Multi-scale modeling and domain partitioning

[0076] First, based on the actual geometric model of the nuclear fusion device or tooling, the overall structure is divided into regions. The core of this step is to establish a multi-scale finite element model: welds and heat-affected zones are modeled using fine meshes to ensure local calculation accuracy; the main structure far from the weld area is modeled using coarse meshes to reduce the overall number of degrees of freedom. Continuous coupling between coarse and fine meshes is achieved using the MPC or Tie method in Abaqus, thus obtaining a multi-scale model that balances computational efficiency and accuracy. The output of this step is a basic finite element analysis model containing global and local scales, providing geometric and mesh support for subsequent heat source loading and physics field calculations.

[0077] Step 2: Welding heat source modeling and heat input definition

[0078] After completing the geometry and mesh generation, the next step is to physically model the energy input of the welding process. The Goldak double-ellipsoidal heat source model is used to describe the heat input distribution, and its dynamic movement and loading along the weld path is controlled by the user subroutine DFLUX written in Fortran. The core technology of this step is to achieve a realistic physical representation of the welding heat source. Its input is the multi-scale geometric model and weld path data from step 1, and its output is the heat input distribution field that varies with time and space, laying the physical driving conditions for subsequent temperature field calculations.

[0079] Step 3: Transient heat conduction and material thermal property modeling

[0080] Based on the heat source model, a transient heat conduction equation for the welding process is established, and temperature-varying material thermophysical parameters, such as density, specific heat capacity, thermal conductivity, and latent heat of phase change, are introduced. The core of this step is to obtain the time-varying temperature field distribution through numerical solution. Its input is the heat input distribution defined in step 2, and the output is the evolution result of the temperature field, while providing boundary conditions and driving forces for subsequent thermo-mechanical coupling analysis.

[0081] Step 4: Thermo-mechanical coupling analysis and stress field solution

[0082] After obtaining the temperature field distribution, a thermo-mechanical coupling calculation is further performed. The core of this step is based on the thermo-elastic-plastic constitutive relationship, considering thermal expansion, material plasticity, and yielding behavior, to obtain the stress and deformation fields that evolve over time. The input is the temperature field from step 3, and the output is the residual stress distribution and transient deformation results, providing a mechanical response basis for the continuous calculation of the weld segment model and subsequent splicing.

[0083] Step 5: Automatic generation of weld segment model and heat history transfer

[0084] For complex long welds, a Python script is used to segment the weld path, generating an independent sub-model for each weld segment and automatically submitting it for calculation. The core technology of this step is the thermal history transfer mechanism: the boundary node temperature obtained from the previous welding simulation is used as the initial thermal boundary condition for the next segment, achieving continuity of the welding thermal history. The input is the temperature and stress results calculated in step 4, and the output is the welding effect data of each sub-model segment, providing distributed simulation units for global stitching.

[0085] Step 6: Stress / deformation result splicing and full structural response reconstruction

[0086] After completing the independent calculations for each weld segment, the results of the sub-models are stitched together into the global structure using methods such as stress projection, mesh interpolation, or thermal field reconstruction. The core of this step is establishing a mapping relationship between the segmented simulations and the overall structural response, ensuring the physical continuity and spatial consistency of the global residual stress field and the welding deformation field. The input is the simulation results of each segment from step 5, and the output is a visualized reconstruction of the global residual stress distribution and overall welding deformation, providing ultimately usable data support for engineering applications.

[0087] Step 7: Python Automatic Control Flow and Extended Interface

[0088] Finally, Python scripts are used to automate the entire process of scheduling and data management, including sub-model construction, heat source path loading, simulation task submission, result extraction, and inter-segment data transmission. The core technology of this step lies in process integration and open interfaces, which not only significantly reduces manual operation and debugging time but also provides extended interfaces for subsequent introduction of optimization algorithms (such as genetic algorithms and Bayesian optimization), process database construction, and intelligent process recommendation. The input is the analysis process of the first six steps, and the output is standardized simulation results and visualized data, ultimately forming a reusable and scalable multi-scale welding simulation and optimization framework.

[0089] The interrelationships of the above steps are explained below:

[0090] Step 1: Multi-scale modeling and structural domain partitioning (as the basis for welding heat source modeling)

[0091] ① Input: Geometric model of the nuclear fusion device and weld layout.

[0092] ② Processing: Fine mesh (weld seam + heat-affected zone) + coarse mesh (non-critical area) + mesh transition coupling.

[0093] ③ Output: Multi-scale finite element model, providing the geometric and mesh basis for heat source loading and subsequent calculations.

[0094] Step 2: Welding heat source modeling and heat input definition (providing driving force for temperature field calculation)

[0095] ① Input: Multi-scale finite element model and weld path data from step 1.

[0096] ② Processing: Goldak double ellipsoidal heat source model + DFLUX subroutine to control the movement of the heat source.

[0097] ③ Output: Spatial-temporal distributed welding heat input field.

[0098] Step 3: Transient heat conduction and material thermophysical property modeling (temperature field as load input for thermo-mechanical coupling)

[0099] ① Input: Thermal input field from step 2 + material thermophysical parameters (varying with temperature).

[0100] ② Processing: Solve the transient heat conduction equation to obtain the temperature distribution over time and space.

[0101] ③ Output: Temperature field evolution results during the welding process.

[0102] Step 4: Thermo-mechanical coupling analysis and stress field solution (as the core result for the piecewise model and subsequent splicing)

[0103] ① Input: Temperature field distribution in step 3.

[0104] ② Treatment: Thermo-elastic-plastic constitutive relations are used, taking into account thermal expansion, plasticity, and yielding behavior.

[0105] ③ Output: Residual stress field and welding deformation field.

[0106] Step 5: Automatic generation of weld segment model and transfer of thermal history (multiple segment results serve as the basis for global stitching)

[0107] ① Input: Temperature / stress results from step 4.

[0108] ② Processing: Python automatically generates sub-models, using the temperature field of the previous segment as the initial condition for the next segment, thus achieving continuity in the heat history transfer.

[0109] ③ Output: Independent simulation results for each weld segment (residual stress, deformation, thermal field).

[0110] Step 6: Stress / deformation results stitching and full structural response reconstruction (results enter automated management and optimization extension)

[0111] Input: Simulation results for each segment in step 5.

[0112] Processing: Projection, interpolation, and thermal field reconstruction to achieve spatial splicing of segmented results.

[0113] Output: Visualized reconstruction of the residual stress field of the entire structure and the overall welding deformation field.

[0114] Step 7: Python Automatic Control Flow and Extended Interface

[0115] ① Input: The calculation process and results of the first 6 steps.

[0116] ② Processing: Python is used to automate the entire process scheduling and data management, with reserved interfaces for optimization algorithms / process databases / intelligent recommendations.

[0117] ③ Output: Standardized simulation results data, visualization reports, and an extensible interface platform.

[0118] The logical relationship between the 7 steps is as follows:

[0119] Geometric modeling (Step 1) → Heat source drive (Step 2) → Temperature field (Step 3) → Stress field (Step 4) → Piecewise continuity (Step 5) → Global reconstruction (Step 6) → Automation and expansion (Step 7).

[0120] The following is a detailed explanation:

[0121] like Figure 1 As shown in the schematic diagram, the multi-scale simulation and optimization method for welding of nuclear fusion devices provided in this embodiment of the invention fully demonstrates the technical path from multi-scale modeling, definition of welding heat sources, calculation of heat conduction and thermo-mechanical coupling, to automatic generation of weld segment models, heat history transfer, result splicing, and fully automated control of the entire process. This flowchart intuitively reflects the input-output relationships and logical connections between each step, indicating that the invention achieves a balance between simulation accuracy and computational efficiency, and possesses good scalability.

[0122] Step 1: Multi-scale modeling and domain partitioning

[0123] Based on the actual geometric model of the nuclear fusion device or tooling, the regions are divided as follows:

[0124] ① Fine mesh modeling (unit size 0.5–2 mm) is used for welds and heat-affected zones;

[0125] ② Structures far from the welding area are modeled using coarse meshes (unit size 5–20 mm);

[0126] ③ Use the MPC or Tie method in Abaqus to achieve multi-scale transition coupling between coarse and fine meshes;

[0127] ④ For complex long welds, further discretize them along the welding path into N physical segments as subsequent segmented simulation units.

[0128] Step 2: Welding heat source modeling and heat input definition

[0129] The welding heat input is defined using the Goldak double ellipsoidal heat source model, and its heat source volume distribution function is:

[0130]

[0131] In the formula,

[0132] Heat flux per unit volume, in W / m³ 3 ;

[0133] Total heat input power, in watts (W).

[0134] The length of the heat source along the x, y, z axes is in meters (m).

[0135] This is the weighting coefficient for the distribution of heat source areas before and after the heat source, and it is dimensionless.

[0136] The user subroutine DFLUX, written in Fortran, controls the heat source to move and load along each weld seam path, thus achieving physical segmentation simulation of the weld seam.

[0137] Step 3: Transient heat conduction and material thermal property modeling

[0138] Temperature field calculation based on transient heat conduction control equations:

[0139]

[0140] In the formula,

[0141] The density of the material is expressed in kg / m³. 3 ;

[0142] This is the specific heat capacity at constant pressure of the material, expressed in J / (kg*K).

[0143] The thermal conductivity of the material is expressed in W / (m*K).

[0144] Temperature, in Kelvin (K).

[0145] This is the heat source term, with units of W / m³. 3 ;

[0146] Temperature-dependent thermophysical properties It can be interpolated and fitted based on the measured material data, and a phase transformation latent heat model can be introduced to handle the austenite / martensite transformation process.

[0147] Step 4: Thermo-mechanical coupling analysis and stress field solution

[0148] After obtaining the temperature field, calculations of thermally induced deformation and stress evolution are performed. A thermo-elastic-plastic constitutive model is used, and its formulas are as follows:

[0149]

[0150] In the formula,

[0151] This is the stress tensor, with units of Pa.

[0152] This is the elastic stiffness tensor, with units of Pa.

[0153] The total strain tensor is dimensionless;

[0154] This is the coefficient of thermal expansion, with units of 1 / K;

[0155] The initial temperature is expressed in Kelvin (K).

[0156] To improve the accuracy of stress prediction during welding, an isotropic strengthening model or an anisotropic yield criterion can be considered to simulate the nonlinear response of materials.

[0157] Step 5: Automatic generation of weld segment model and heat history transfer

[0158] ① Each weld segment is automatically generated as an independent sub-model (controlled by Python);

[0159] ② The boundary node temperature calculated in the previous segment is used as the thermal boundary of the next segment, and the thermal history is continuously transferred through *Temperature or *MAP;

[0160] ③ After each simulation is completed, the residual stress and deformation results are saved for subsequent splicing or reconstruction.

[0161] like Figure 2 The figure shows a schematic diagram of the local residual stress distribution in the segmented welding area provided by an embodiment of the present invention. This figure illustrates the residual stress field distribution in the local weld area after multi-segment welding simulation using the method of the present invention. As can be seen from the figure, the stress concentration area can be clearly captured, and the residual stress gradually decreases in the weld and heat-affected zone, verifying the accuracy and physical rationality of the method of the present invention in stress prediction.

[0162] like Figure 3The diagram shown is a schematic representation of the local welding deformation distribution in the segmented welding zone provided in an embodiment of the present invention. This diagram reflects the local structural deformation characteristics caused by heat input and constraint conditions during the welding process. The deformation is mainly concentrated near the weld and gradually decreases towards the surrounding structure. This result allows for a direct assessment of the impact of heat input on the overall welding deformation, thus providing a clear basis for process optimization.

[0163] In this step, a Python script is used to segment the weld path. In this embodiment, the following two methods are used:

[0164] Example A: Multi-segment simulation method for welds based on inter-segment error self-calibration

[0165] This embodiment is a key step in the multi-scale segmented simulation framework of the present invention. By automatically evaluating the continuity of the thermal and stress fields of two adjacent weld segments, and automatically triggering segment length adjustment, mesh refinement or boundary condition correction based on the error results, it ensures that the segmented calculation results of the entire weld are consistent with the solution accuracy of the full model.

[0166] Step 151: Boundary field extraction after the previous weld simulation is completed

[0167] Regarding the first For the segmented welding model, after the thermo-mechanical coupling analysis is completed, the odbAccess module in Python is used to extract the following data from the weld end section of the .odb file:

[0168] Temperature field:

[0169]

[0170] Stress field:

[0171]

[0172] in, This represents the number of nodes in this cross section.

[0173] Save the extracted file as:

[0174] S_i_boundary_T.dat

[0175] S_i_boundary_S.dat

[0176] Step 152: Map the boundary data to the next segment of the initial model.

[0177] Construct the first After creating the segment model, the mapping module in Python is called to map the boundary temperature and stress of the previous segment to the starting section of the next segment model:

[0178]

[0179] Mapping methods can include nearest neighbor interpolation, inverse distance weighting (IDW), or two-dimensional cubic interpolation.

[0180] After mapping is completed, the initial boundary condition data for the next segment is formed.

[0181] Step 153: Calculate the errors in the inter-segment thermal and stress fields.

[0182] Compare the differences between the mapping values ​​at the end of the previous paragraph and the next paragraph:

[0183] Thermal field error calculation:

[0184]

[0185] Stress field error calculation:

[0186]

[0187] Step 154: Perform self-calibration judgment by comparing with the error threshold.

[0188] Introduce two user-specified error thresholds:

[0189] Thermal field error threshold: (Generally, 5% is taken)

[0190] Stress field error threshold: (Generally, 5% is taken)

[0191] The judgment condition is:

[0192] Heat field assessment:

[0193]

[0194] Stress field determination:

[0195]

[0196] If both conditions are met, the physical continuity between segments is considered to meet the requirements, and the simulation process proceeds to the next segment.

[0197] Step 155: Three Automatic Calibration Measures When Requirements Are Not Met

[0198] like: or

[0199] This invention automatically enters the self-calibration process, as follows:

[0200] (1) Adaptively shorten the length of the next segment

[0201] Let the original segment length be... Then adjust to:

[0202]

[0203] in, The value range is 0.5~1, and the length of the next segment is automatically adjusted to reduce the cross-segment attenuation error of the thermal field.

[0204] (2) Automatically trigger local mesh refinement

[0205] If the error originates from a coarse mesh, the Python-CAE interface is automatically invoked to adjust the mesh size near the initial section to a smaller size and automatically generate a new .inp file.

[0206] (3) Automatically construct the transition temperature field

[0207] When mapping errors arise from coordinate distortion or changes in cross-sectional shape, a linear transition field is constructed:

[0208]

[0209] in, The smoothing factor (secondary optimization parameter) has a value range of 0 to 1. Constructing a transition field can avoid numerical instability caused by "jumping temperature field".

[0210] Step 156: After self-calibration is complete, regenerate the next segment of the model and repeat the judgment.

[0211] After calibration, the Python script automatically regenerates the model input file, initial boundary conditions, and heat source path control file, and recalculates the error until the error judgment conditions are met.

[0212] Finally, we move on to the next stage of the solution process.

[0213] Through this embodiment, the present invention can achieve continuity of temperature field across segments, smooth transition of stress field, and segmented calculation results that approximate the accuracy of the full model solution, while avoiding the problem of numerical offset accumulation caused by segmentation.

[0214] Example B: Adaptive Weld Segment Length Algorithm

[0215] In this embodiment, to improve the accuracy and efficiency of segmented welding simulation, the present invention automatically determines the segment length of each weld based on the geometric features of the weld path, the change in structural thickness, and the distribution of fixture constraints, thereby achieving adaptive segmentation of the weld path.

[0216] Step 251: Extract weld path data and calculate geometric features.

[0217] Export the weld centerline point sequence from the CAD model:

[0218]

[0219] Calculate the local curvature of the weld using the three-point curvature formula:

[0220]

[0221] Curvature represents the degree of local bending of the weld. The greater the curvature, the more obvious the change in the direction of heat input, and the shorter the segments must be to ensure the accuracy of the heat source path.

[0222] Step 252: Extract the structural thickness change rate

[0223] Read the structural thickness near the weld from the model. And calculate the thickness gradient:

[0224]

[0225] in, For thickness, This indicates the path location.

[0226] The greater the thickness variation, the more pronounced the difference in thermal diffusion behavior, requiring denser segmentation to capture the heat transfer gradient.

[0227] Step 253: Collect fixture constraint distribution information

[0228] Obtain the constraint distribution function of the weld region using the fixture model:

[0229]

[0230] The denser the fixtures and the stronger the constraints, the more abrupt the changes in welding stress, and the shorter the number of segments should be.

[0231] Step 254: Adaptive Segment Length Formula

[0232] The formula for calculating segment length is:

[0233]

[0234] in,

[0235] For the weld in position The required segment length, in mm;

[0236] Set the maximum segment length for the user, in mm;

[0237] The curvature of the weld is expressed in 1 / mm.

[0238] This is the thickness change rate, in units of 1 / mm;

[0239] The constraint strength function;

[0240] , , This is a weighting coefficient, which can be qualitatively set according to the "degree of influence on simulation results". The default value is set to 1, and it can be obtained through training from the historical process database.

[0241] This formula guarantees that: large weld bends will automatically shorten the segment length; large thickness variations will automatically increase the segment density; and high constraint strength will automatically enhance local precision.

[0242] Step 255: Automatically generate variable segmented sequences

[0243] The Python script is based on the calculation The values ​​automatically generate the start and end positions of each segment, forming:

[0244] segment_1.json

[0245] segment_2.json ...

[0246] For use in subsequent modeling and heat source loading.

[0247] Step 256: Link with the main simulation process

[0248] The path file for each segment is passed to a Fortran subroutine (DFLUX) to control the movement of the heat source for that segment, enabling accurate simulation of heat input.

[0249] When combined with the error calibration mechanism of Example A, if a certain segment of error exceeds the limit, the present invention can further automatically shorten the length of that segment to improve simulation accuracy.

[0250] By using the above adaptive segmentation method, this invention significantly improves the shortcomings of the traditional fixed segmentation method, enabling the model to have higher local simulation accuracy in areas with large weld curvature, significant thickness variation, and dense constraints, while reducing the number of unnecessary segments and improving the overall simulation efficiency.

[0251] Step 6: Stress / deformation result splicing and full structural response reconstruction

[0252] The full structural welding effect can be reproduced using one of the following methods:

[0253] ① Stress projection method: Projecting the stress field of the sub-model onto the global structure;

[0254] ② Temperature back-calculation method: Construct a continuous heat input field and recalculate the residual stress on the overall structure;

[0255] ③ Linear superposition method: applicable when there is no strong overlap between heat-affected zones of sections.

[0256] Step 7: Python Automatic Control Flow and Extended Interface

[0257] Python scripts are used for:

[0258] Automatically generate the mesh and heat source path for each segment of the model;

[0259] Controlling Fortran subroutine calls;

[0260] Automatically submit simulation tasks and manage the solution order;

[0261] Extract simulation results and generate visualized data;

[0262] Reserve interfaces for future optimization algorithms (such as genetic algorithms and Bayesian optimization).

[0263] In practical implementation, this invention can introduce a parallel computing architecture to distribute the simulation tasks of each weld segment to multi-core processors or distributed computing nodes for parallel solving, thereby further shortening the total time required for the overall welding simulation process and improving the engineering efficiency of large-scale assembly weld evaluation.

[0264] This invention can automatically generate optimized welding sequence and path layout by integrating intelligent weld path planning algorithms (such as heuristic search or reinforcement learning methods) to reduce heat input overlap, suppress deformation accumulation, and complete the integrated closed loop of path design-verification-optimization in conjunction with the simulation system.

[0265] This invention can construct a welding process database, store historical simulation and experimental results in a structured data platform, and train a prediction model using machine learning methods. This allows for the rapid evaluation of residual stress and welding deformation trends of new process combinations without the need for full-process simulation, thus achieving a coupled intelligent process recommendation system of "simulation-prediction-optimization".

[0266] This invention can construct an inversion optimization module for simulation calibration by introducing measured data obtained by digital image correlation (DIC) technology or X-ray residual stress measurement technology, so as to realize the inverse calibration of key input parameters such as material constitutive parameters, heat source efficiency, and boundary conditions, and significantly improve the reliability and engineering adaptability of simulation prediction.

[0267] This invention can couple with a structural strength verification module to input the welding residual stress field as a load into the fatigue life assessment and failure analysis process, thereby realizing an integrated virtual evaluation system for the entire process from welding manufacturing to service performance analysis.

[0268] The splicing technology involved in step 6 is specifically implemented as in Example C;

[0269] Example C: Weighted splicing method for stress and deformation fields of multi-segment welds based on temperature gradient

[0270] After multi-segment welding simulation, this invention proposes a weighted splicing model based on temperature gradient to obtain the integrated residual stress field and deformation field of the entire weld. This model smoothly integrates the calculation results of adjacent weld segments, avoiding the problems of field abrupt changes, stress discontinuity and numerical discontinuity that occur in traditional segment splicing.

[0271] Step 61: Extract field variables from adjacent weld segments

[0272] For the Section and the From the simulation results, extract the nodes of the splicing region:

[0273] temperature: ,

[0274] stress: ,

[0275] Displacement: ,

[0276] Step 62: Calculate the nodal temperature gradient

[0277] The local temperature gradient is calculated based on the temperature field, and the temperature gradient reflects the degree of change in the heat-affected zone during welding.

[0278]

[0279] Step 63: Construct a weighting function based on the temperature gradient

[0280] Define weights:

[0281]

[0282] in, This is a smoothing factor, ranging from 0.5 to 2, used to control the smoothness of the stitching. A larger temperature gradient results in a smaller weight, indicating that the field data is more dependent on the other data segment; a smaller temperature gradient results in a weight close to 1, indicating that the two segments contribute similarly.

[0283] Step 64: Weighted fusion of field variables

[0284] Stress, displacement, and other field quantities are spliced ​​together in the following form:

[0285]

[0286]

[0287] This ensures that the field data in the splicing area is physically continuous and numerically smooth.

[0288] Step 65: Stress and deformation distribution forming the integral weld

[0289] The fused splicing region results are mapped to the global model to obtain the residual stress field and deformation field of the complete weld, providing unified data for subsequent structural analysis or process optimization.

[0290] This method can eliminate abrupt changes and discontinuities caused by segmented calculations, improve the smoothness and reliability of the overall welding field data, and ensure that the multi-segment simulation results are more consistent with the actual welding behavior. At the same time, it can be combined with the methods involved in the above embodiments A and B to form a three-dimensional high-precision segmented simulation system.

[0291] To further illustrate the technical solution of the present invention, a specific embodiment is provided in conjunction with a typical application scenario. This embodiment is used for modeling and simulating the welding process of a typical large cylindrical structure in a nuclear fusion device, illustrating the feasibility and steps of the present invention in engineering applications.

[0292] I. Three-dimensional geometric modeling and weld segment division

[0293] (1) Structural modeling

[0294] Use CAD software (such as SolidWorks) to create a complete welded structure, including the upper and lower cylinder sections to be welded, the weld zone, and the flange connection area. Divide the weld into N segments according to length, denoted as... The length of each segment is set to This length is typically 2 to 3 times the width of the heat-affected zone of the weld.

[0295] (2) Export weld path data

[0296] Extract the weld centerline path point sequence from the CAD file and save it in a standard format (such as CSV or JSON). An example of the format is shown below:

[0297] segment_id, point_index, x(mm), y(mm), z(mm)

[0298] S1, 0, 0.0, 0.0, 0.0

[0299] S1, 1, 5.0, 0.0, 0.0 ...

[0300] S2, 0, ...

[0301] II. Automatic Modeling and Segment Model Generation using Python Scripts

[0302] (1) Automatic sub-model generation (Python controls Abaqus / CAE)

[0303] Write a Python script named build_segment_model.py with the following functionality:

[0304] Read the path data file;

[0305] For each weld segment Create an Abaqus model;

[0306] Extract the center path of the current segment and generate a local modeling region (which can be a rectangular block, a sector, etc.);

[0307] Set a fine mesh generation strategy (C3D8RT cells, mesh size 0.5 to 1.0 mm).

[0308] Set the boundary region as a thermally conductive or rigidly constrained boundary.

[0309] Automatically generate .cae and .inp files.

[0310] (2) Example of segment model file naming conventions:

[0311] vacuum_weld_S1.inp

[0312] vacuum_weld_S2.inp ...

[0313] III. Defining the Welding Heat Source (DFLUX) using a Fortran subroutine

[0314] (1) Write the DFLUX subroutine: weld_dflux.for

[0315] Implement the volumetric heat flux distribution function for a Goldak double-ellipsoidal heat source. The Fortran subroutine needs to receive the current location, parameters, and time of the heat source, and return the unit volume heat input at each integration point. .

[0316] The core computing components are illustrated below:

[0317] subroutine dflux(...)

[0318] ! Input parameters

[0319] ! x(1:3): Spatial coordinates of the current integration point [mm]

[0320] ! time: Current simulation time point [s]

[0321] ! coordinates of current heat source center: xh, yh, zh

[0322] ! heat input Q, a, b, c, f_front, f_rear

[0323] ! Calculate the relative distance between the current location and the center of the heat source.

[0324] dx = x(1) - xh

[0325] dy = x(2) - yh

[0326] dz = x(3) - zh

[0327] Determine whether it is in the front or back region, and calculate q(x,y,z).

[0328] if (dx>= 0) then

[0329] q = (6 * sqrt(3) * f_front * Q) / (pi * sqrt(pi) * a * b * c) *

[0330] exp(-3 * ((dx / a)**2 + (dy / b)**2 + (dz / c)**2))

[0331] else

[0332] q = (6 * sqrt(3) * f_rear * Q) / (pi * sqrt(pi) * a * b * c) *

[0333] exp(-3 * ((dx / a)**2 + (dy / b)**2 + (dz / c)**2))

[0334] endif

[0335] dflux = q

[0336] end subroutine

[0337] (2) Key input parameter description:

[0338] Total heat input power, in watts (W).

[0339] The length of the heat source along the x, y, z axes is in meters (m).

[0340] , which is the weighting coefficient for the distribution in the front / back zone of the heat source, and is dimensionless;

[0341] This represents the current location of the heat source center, expressed in meters (m).

[0342] (3) Integration with Python

[0343] Python writes the coordinate sequence of each path segment into the file weld_path_Si.dat for DFLUX to read. DFLUX indexes the path point position based on the current time point, realizing the movement of the heat source over time.

[0344] IV. Simulation Scheduling and Thermal History Transfer (Python Control)

[0345] (1) Thermal boundary derivation and transfer

[0346] After each simulation is completed, Python uses the Abaqus odbAccess module to read the boundary node temperature and saves it as S1_boundary_temperature.dat, in the following format:

[0347] node_id, T(K)

[0348] 10123, 1180.5

[0349] 10124, 1179.8 ...

[0350] (2) Next segment of model thermal boundary assignment

[0351] Python inserts the following when building the next model.inp file:

[0352] *Temperature, file=S1_boundary_temperature.dat

[0353] Alternatively, you can use *MAP SOLUTION to directly map the thermal field boundary from the previous .odb file.

[0354] (3) Simulation batch processing scheduling script illustration

[0355] for i in range(1, N+1):

[0356] segment_name = f"S{i}"

[0357] # 1. Constructing a segment model

[0358] build_segment_model(segment_name)

[0359] # 2. Solve using Abaqus

[0360] os.system(f"abaqus job={segment_name} user=weld_dflux.f90 cpus=8")

[0361] # 3. Extract thermal boundary data

[0362] extract_temperature_boundary(segment_name)

[0363] # 4. Prepare the next set of boundary conditions

[0364] if i <N:

[0365] transfer_temperature_to_next(segment_name, f"S{i+1}")

[0366] V. Post-processing and result stitching

[0367] (1) Extracting result fields

[0368] Python automatically extracts the following from each .odb file:

[0369] Residual stresses at nodes (S11, S22, S33)

[0370] Total deformation of the weld area (U1, U2, U3)

[0371] Width of the local heat-affected zone (analyzed by temperature contour lines)

[0372] (2) splicing strategy

[0373] Map the segment results to a unified virtual global grid (using trilinear interpolation or nearest neighbor projection).

[0374] If a complete structural coupling recalculation is required, the temperature history can be interpolated to the coarse mesh structural model, and the global residual stress can be solved once again.

[0375] In summary, the embodiments of the present invention have the following characteristics:

[0376] ① Weld path segmentation activation and multi-segment sub-model collaborative simulation mechanism

[0377] Discretizing the entire complex weld path into multiple physical segments and constructing an independent sub-model for each segment, and then applying heat sources segment by segment and simulating and solving segment by segment, significantly reduces the degrees of freedom and resource consumption of the entire model while maintaining high simulation accuracy locally. This is one of the core ideas of this invention.

[0378] ② Dynamic path loading method controlled by Fortran heat source subroutine

[0379] By embedding the Abaqus heat source control flow into a Fortran subroutine (DFLUX) and combining it with the weld path point sequence passed in from Python, the welding heat source is dynamically moved and loaded along the weld segment over time. This achieves physical modeling of the real welding process rather than temperature-assumed loading, significantly improving the accuracy of temperature field simulation.

[0380] ③ Multi-scale mesh coupling modeling and automatic sub-model generation technology

[0381] The modeling strategy of "fine mesh in critical areas + coarse mesh in non-critical areas" is adopted, combined with the automatic sub-model construction and mesh generation logic under Python control, to ensure that the modeling process is repeatable, automated, and highly adaptable to the structure, thereby reducing human error and development costs.

[0382] ④ The continuous transfer mechanism of thermal boundary conditions between sections

[0383] By extracting the boundary node temperature field from the previous simulation results and using it as the initial thermal boundary condition for the next model, the physical continuity of the thermal history is maintained, thus solving the problems of thermal field fragmentation and large prediction errors in traditional segmented models.

[0384] ⑤ Python-controlled process automation and simulation scheduling mechanism

[0385] This method automatically schedules all processes, including modeling, path loading, parameter configuration, Abaqus calls, result extraction, and inter-segment data transfer, through Python scripts. It supports batch weld segment simulation, parallel computing, and efficient data management, and is a key support for industrial deployment.

[0386] ⑥ Method for splicing multi-segment welding results and reconstructing the global field

[0387] By using techniques such as node mapping, mesh interpolation, or thermal field reconstruction, the simulation results of multiple weld segments are spliced ​​and fused in space to achieve a visual reconstruction of the residual stress field and welding deformation field of the entire structure, which is convenient for subsequent assembly deviation assessment and structural optimization.

[0388] Based on the aforementioned key technical approaches, the embodiments of the present invention specifically include the following contents.

[0389] A multi-segment activated welding simulation method that physically segments the weld path and independently loads heat sources and solves temperature and stress fields in each segment's local model; a method for controlling the movement trajectory of heat sources using Fortran subroutines, where the heat source trajectory is driven by an external path file (e.g., generated by Python) and dynamically changes; an automatic transfer mechanism for inter-segment thermal boundary conditions, including node temperature export and assignment or mapping of the next segment's thermal boundary; a Python workflow system for automating sub-model modeling and simulation scheduling, covering modeling, heat source configuration, solving, boundary transfer, and post-processing; a method for stitching together multi-segment simulation results to form the residual stress and deformation field of the entire structure, including mapping, interpolation, and thermal field reconstruction; and a collaborative framework for multi-scale finite element modeling and welding simulation applicable to large-scale complex structures in nuclear fusion devices.

[0390] Any combination or integrated application scenario of the above points, especially extended functions such as welding path optimization, fixture effect analysis, and heat input sensitivity analysis.

[0391] In another aspect, the present invention also discloses a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method described above.

[0392] In another aspect, the present invention also discloses a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method described above.

[0393] It is understood that the systems, devices, and storage media provided in the embodiments of the present invention correspond to the methods provided in the embodiments of the present invention, and the explanations, examples, and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

[0394] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0395] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0396] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0397] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-scale simulation and optimization method for welding in a nuclear fusion device, characterized in that, Includes the following steps, S1. Divide the overall structure into regions based on the actual geometric model of the nuclear fusion device or tooling; S2. After completing the geometry and mesh generation, perform physical modeling of the energy input of the welding process. The input is a multi-scale geometric model and weld path data, and the output is the heat input distribution field that varies with time and space. S3. Based on the heat source model, establish the transient heat conduction equation during the welding process and introduce material thermal property parameters that vary with temperature; The input is the heat input distribution defined by S2, and the output is the evolution result of the temperature field; S4. After obtaining the temperature field distribution, perform thermo-mechanical coupling calculations; the input is the temperature field from S3, and the output is the residual stress distribution and transient deformation results. S5. For complex long welds, a Python script is used to segment the weld path. Each weld segment generates an independent sub-model and is automatically submitted for calculation. The input is the temperature and stress results calculated in S4, and the output is the welding effect data of each segment sub-model. The segmentation of the weld path is implemented using Python scripts, including the following two methods: A. By automatically assessing the continuity of the thermal and stress fields between two adjacent weld segments, and automatically triggering segment length adjustment, mesh refinement, or boundary condition correction based on the error results; B. Based on the geometric features of the weld path, the variation of structural thickness, and the distribution of fixture constraints, the segment length of each weld segment is automatically determined. S6. After completing the independent calculation of each weld segment, the results of the sub-model are stitched into the global structure through stress projection, mesh interpolation or thermal field reconstruction methods; the input is the simulation results of each segment in S5, and the output is the visualized reconstruction results of the global residual stress distribution and the overall welding deformation. The splicing steps include, To obtain the integrated residual stress field and deformation field of the entire weld, a weighted splicing model based on temperature gradient is adopted to smoothly merge the calculation results of adjacent weld segments. The temperature gradient-based weighted splicing model extracts the temperature, stress, and displacement field variables of the splicing area of ​​adjacent weld segments and calculates the local temperature gradient. It then constructs a weighting function that dynamically changes with the temperature gradient and weights and fuses the field variables of adjacent segments to achieve physical continuity and numerical smooth splicing of stress and deformation fields of multiple weld segments. S7. Use Python scripts to automate the entire process of scheduling and data management from S1 to S6.

2. The welding multi-scale simulation and optimization method for nuclear fusion devices according to claim 1, characterized in that: S1 specifically includes, The weld and heat-affected zone are modeled using a fine mesh with a unit size of 0.5–2 mm; Structures far from the welding area are modeled using coarse meshes with element sizes of 5–20 mm. Use the MPC or Tie methods in Abaqus to achieve multi-scale transition coupling between coarse and fine meshes; For complex long welds, they are discretized into N physical segments along the welding path, which serve as the subsequent segmented simulation units.

3. The welding multi-scale simulation and optimization method for nuclear fusion devices according to claim 2, characterized in that: Step S2 specifically includes, The welding heat input is defined using the Goldak double ellipsoidal heat source model, and its heat source volume distribution function is: In the formula, Heat flux per unit volume, in W / m³ 3 ; Total heat input power, in watts (W). The length of the heat source along the x, y, z axes is in meters (m). , which is the weighting coefficient for the distribution in the front / back zone of the heat source, and is dimensionless; The user subroutine DFLUX, written in Fortran, controls the heat source to move and load along each weld seam path, thus achieving physical segmentation simulation of the weld seam.

4. The welding multi-scale simulation and optimization method for nuclear fusion devices according to claim 3, characterized in that: S3 specifically includes, Temperature field calculation based on transient heat conduction control equations: In the formula, The density of the material is expressed in kg / m³. 3 ; This is the specific heat capacity at constant pressure of the material, expressed in J / (kg*K). The thermal conductivity of the material is expressed in W / (m*K). Temperature, in Kelvin (K). This is the heat source term, with units of W / m³. 3 ; Temperature-dependent thermophysical properties The material was fitted by interpolation based on measured data, and a latent heat model of phase transformation was introduced to handle the austenite / martensite transformation process.

5. The welding multi-scale simulation and optimization method for nuclear fusion devices according to claim 4, characterized in that: S4 specifically includes, After obtaining the temperature field, calculations of thermally induced deformation and stress evolution are performed; a thermo-elastic-plastic constitutive model is adopted, and its formula is as follows: In the formula, This is the stress tensor, with units of Pa. This is the elastic stiffness tensor, with units of Pa. The total strain tensor is dimensionless; This is the coefficient of thermal expansion, with units of 1 / K; The initial temperature is expressed in Kelvin (K).

6. The welding multi-scale simulation and optimization method for nuclear fusion devices according to claim 5, characterized in that: S5 specifically includes, Each weld segment is automatically generated as an independent sub-model; The boundary node temperature calculated in the previous segment is used as the thermal boundary in the next segment, and the continuous transfer of thermal history is achieved through *Temperature or *MAP. After each simulation is completed, the residual stress and deformation results are saved for subsequent splicing or reconstruction.

7. The welding multi-scale simulation and optimization method for nuclear fusion devices according to claim 6, characterized in that: S6 specifically includes, The full structural welding effect can be reproduced using one of the following methods: Stress projection method: Projecting the stress field of the sub-model onto the global structure; Temperature-based reverse calculation method: Construct a continuous heat input field and recalculate the residual stress on the overall structure; Linear superposition method: applicable when there is no strong overlap between heat-affected zones of different sections.

8. The welding multi-scale simulation and optimization method for nuclear fusion devices according to claim 7, characterized in that: S7 specifically includes execution via Python scripts: Automatically generate the mesh and heat source path for each segment of the model; Controlling Fortran subroutine calls; Automatically submit simulation tasks and manage the solution order; Extract simulation results and generate visualized data; An interface is reserved for future optimization algorithms.

Images