92 results about "Geometric design" patented technology

The invention discloses a complex part molded line optimization design method based on multi-field node feature weighted aggregation. Comprising the following steps: firstly, performing finite element batch simulation according to a geometric design variable, a temperature and a contact surface friction coefficient of a to-be-optimized complex part molded line to obtain a training data set; then, constructing a depth map sampling aggregation neural network fusing space and temperature attention, then training the depth map sampling aggregation neural network fusing space and temperature attention by using the training data set, and obtaining a stress prediction model after training is completed; and finally, in combination with the stress prediction model, performing iterative optimization by taking minimization of the maximum equivalent stress of the complex part as an optimization target, and obtaining the optimal geometric parameters of the molded line of the current complex part in the target temperature interval. According to the method, rapid reconstruction of the stress distribution state of the complex part is achieved, the optimization speed is increased, meanwhile, the precision of an optimization result is fully guaranteed, and meanwhile calculation resources are greatly saved.

The invention discloses a full-variable optimized B-spline geometric distance fitting method, belongs to the field of computer-aided geometric design, and is suitable for solving the problem of high-precision modeling of complex curves in industrial design and precision manufacturing scenes. According to the method, variables are initialized for a set of given data points; constructing a target function of geometric distance fitting, solving update quantities of control points and nodes through a Gauss-Newton algorithm, and normalizing node vectors; approaching projection points of the data points by using first-order approximation to update parameters; and checking a convergence index of alternate iteration to obtain a final B-spline curve. Compared with the prior art, the method has the advantages that the fitting precision is remarkably improved, the problem that in the prior art, a dependent variable coupling relation is complex and global optimization is difficult is solved, and particularly when data points containing sharp features are processed, a B-spline curve with heavy nodes can be automatically generated, geometric features can be better described, and fitting errors are reduced.

A vehicle is controlled for travel on a road having a geometric design defined by one or a combination of a docking line, a cross-section, and a cross-section of the road such that different values of parameters of the geometric design of the road, traffic on the road, traffic rules for traffic flow on the road define different traffic scenarios. By relaxing configuration parameters of a real-world scene and tightening corresponding limiting parameters, a mixed integer non-convex constraint optimization problem for a current real-world traffic scene is converted into a mixed integer convex optimization problem for approximate representation of the real-world traffic scene, so that a carrier is controlled. A mixed integer convex optimization problem for the converted approximate representation of the real-world traffic scene is solved to generate current control commands for controlling one or more actuators of the vehicle.

The invention discloses a physical perception-based generative three-dimensional reverse design method and system, which realize efficient exploration and optimization in a three-dimensional design space by constructing a unified physical-geometric potential representation and combining a two-stage optimization strategy. According to the method, shape and physical field data combined training is utilized to obtain potential representation, so that a geometric structure and physical constraints are compactly represented; on the basis, two stages of physical perception optimization are adopted: in the first stage, global exploration is performed in a potential manifold through a gradient-guided diffusion model, and an initial grid is generated; and in the second stage, the initial grid is locally optimized by utilizing target-driven topology maintenance optimization, so that the initial grid gradually approaches the target performance requirement. According to the method, the high-fidelity three-dimensional geometric design can be directly generated, the target performance and the result diversity of the design are remarkably improved, the method can be widely applied to aerospace, energy equipment and complex engineering structure design, and a new technical approach is provided for efficient implementation of the three-dimensional intelligent design.

The disclosed method includes: determining a subsurface space of interest; arranging a source sensor array relative to a receiver sensor array of a Surface Distributed Acoustic Sensing (S-DAS); transmitting, from the source sensor array, a first set of seismic signals indicating baseline measurements; receiving, using the receiver sensor array, the baseline measurements and thereby generate baseline data; injecting fluid into the subsurface space of interest; transmitting, using the source sensor array, a second set of seismic signals indicating fluid plume measurements; receiving, using the receiver sensor array, the fluid plume measurements and thereby generate fluid plume data; determining, based on the baseline data and the fluid plume data, time-shift data and thereby generate fluid plume extent data; formatting the fluid plume extent data into data values that characterize an extent or rate of spread of the fluid plume; and visualizing the extent or rate of spread of the fluid plume.

The invention provides a finite element analysis method and system for geometric design of a flow regulating valve, and relates to the technical field of data processing.The method comprises the steps that 1, spatial geometric feature analysis is conducted on each data point in a discrete contour coordinate dataset, the principal curvature value of each data point on the sealing surface of a metal diaphragm is solved, and the principal curvature value of each data point is calculated; determining a point with the maximum average curvature value on the sealing surface of the metal diaphragm as a core morphology reference point; step 2, establishing a local coordinate system by taking the core morphology reference point as an original point, screening four sealing state points from the discrete contour coordinate data set, solving a corresponding space plane equation, and calculating a space intersection line between the space plane and the sealing surface of the metal diaphragm; according to the method, through accurate positioning of the core reference point, definition of the analysis sub-domain, grid division, mechanical solution, leakage rate calculation and parameter optimization, the dynamic particulate matter problem is effectively improved, and the sealing performance and the micro flow adjusting precision are improved.

The disclosed methods include: determining a central reference point for fluid migration associated with the stored subsurface fluid; generating a radial coverage detection space extending from the central reference point for the fluid migration associated with the stored subsurface fluid; initializing a subsurface computing model for the stored subsurface fluid; overlaying the radial coverage detection space or detection system over the subsurface model for the stored subsurface fluid; execute one or more simulations by adapting or varying the one or more parameters of the subsurface computing model based on fluid plume data or fluid pressure data associated with the stored subsurface fluid; generating, based on the fluid plume data or fluid pressure data, spatial or temporal fluid distribution data associated with the stored subsurface fluid; and generating, based on the spatial or temporal fluid distribution data, PP data and PS data associated with the stored subsurface fluid.

The invention discloses a microwave device multi-physics field optimization method taking electromagnetism as a center based on a feature assistance and proxy model. The method comprises the steps of setting initial values of geometric design variables and multi-physics field design variables; the multi-physical field response with electromagnetism as the core responds at the solution point {xg, xm}, multiple pieces of parallel multi-physical field data are generated, and f represents the frequency; according to the multi-physics field data, constructing a feature-assisted multi-physics field agent model based on a neural network transfer function, and calculating a gradient optimization derivative of the feature-assisted multi-physics field agent model; and on the basis of the gradient optimization derivative and the optimization feature auxiliary multi-physics field agent model, obtaining an optimal solution of the variable. According to the method, the robustness and optimization design efficiency of a microwave device multi-physics field optimization algorithm taking electromagnetism as a core are improved.

The invention relates to a spring stiffness characteristic simulation optimization method and system. The spring stiffness characteristic simulation optimization method comprises the steps of obtaining first stiffness characteristic curve data of a spring through multi-physical field simulation, and extracting and obtaining a first physical morphological characteristic of the spring based on the first stiffness characteristic curve data; establishing a prediction model between the spring input parameters and the first physical morphological characteristics; in robust design optimization iteration, according to the geometric design parameters and the uncertainty parameters and on the basis of the prediction model, corresponding second physical morphological characteristics are obtained, and on the basis of the second physical morphological characteristics, second rigidity characteristic curve data of the spring are obtained through reconstruction; and on the basis of statistical analysis of the second stiffness characteristic curve data, a robust performance evaluation result of the design scheme under the influence of uncertainty is obtained, and according to the robust performance evaluation result, the geometric design parameters are iteratively adjusted until the spring design scheme meeting the robust requirement is obtained.

The invention discloses an overhead working truck bed-shaped arm section optimization method and system considering stability and rigidity. The method comprises the steps that geometric design parameters of all polygonal sections of the box-shaped arm are set, and the geometric design parameters of all the polygonal sections jointly form an adjustable geometric design space; establishing a three-dimensional finite element analysis model of the box-shaped arm, and endowing the three-dimensional finite element analysis model with material attributes; according to the actual use requirement of the box-shaped arm, the analysis working condition of the three-dimensional finite element sub-model is defined as the most stringent working condition combination; and setting an optimization target, executing a multi-target optimization algorithm and solving to obtain geometric design parameters of each polygon section after optimization. According to the method, through scientific multi-objective optimization simulation, the coupling and restricting relation between two core performance indexes of local stability and deflection (rigidity) in boom design is systematically solved, so that the section shape with the optimal comprehensive performance under the given working condition is obtained.

The invention discloses an intersection geometric design and signal control collaborative optimization method based on simulation optimization, and the method comprises the steps: building an intersection optimization model and a simulation model, enabling the input of the intersection optimization model to be intersection traffic demands and geometric conditions, and enabling the output of the intersection optimization model to be intersection geometric design and signal control related decision variables; the output of the simulation model is an operation evaluation index of the corresponding scheme; wherein the simulation model automatically adjusts a decision variable of an intersection design scheme into a geometric structure and a signal timing parameter in the simulation model, simulation is operated to obtain an optimized target value, the optimized target value serves as an adaptive value of a particle swarm optimization algorithm to update the decision variable of the current scheme, and the updated variable value is returned to the intersection optimization model. And scheme transmission and optimization are realized. According to the method, the diversity of simulation output results is fully exerted, the optimization target can be freely replaced on the basis of not modifying the upper layer model, and the flexibility and optimization efficiency of the intersection optimization model are improved.

A computer-implemented method for systematic configuration, assessment and optimization of a thermal transfer device comprising a plurality of thermal transfer elements arranged in a structured way, wherein each thermal transfer element extends along an axial path, the method including the steps of: starting with an initial geometric design of the plurality of thermal transfer elements; using an optimization framework with a computational model for performing computerized simulations of fluid dynamics and heat transfer at least around said thermal transfer elements during intended operation of the thermal transfer device, in order to perform geometric design optimization of each thermal transfer element based on an optimization algorithm targeting predetermined criteria, wherein the optimization framework is configured to iteratively refine the geometric design of each individual thermal transfer element; and wherein the geometric design of multiple thermal transfer elements are allowed to be varied independently of each other during optimization.

The invention belongs to the technical field of data processing, and provides a design parameter extraction method and system based on geometric features, a geometric model is disassembled in combination with the geometric features, all components in the model and the mutual relation of the components can be more accurately understood and analyzed, and the accuracy of data processing is improved; the geometric design parameters are set according to the geometric feature types, so that the accuracy, reliability and scientificity of the design parameters are ensured, and meanwhile, the efficiency of setting the design parameters can be remarkably improved; the integrity of the geometric model is verified, so that the reasonability of the constructed design parameters can be ensured, and geometric loss cannot be caused; by optimizing the design parameters, the total quantity of the design parameters of the complete geometric model can be effectively reduced, the complexity and the data volume of the model can be remarkably reduced, the maintainability and the expandability of the relevant performance corresponding to the design parameters can be improved, and the cost required for performance evaluation in the later period can be further reduced.

The invention discloses a low-temperature deep-well pump impeller optimization method based on a PSO algorithm. The low-temperature deep-well pump impeller optimization method comprises the following steps that parametric modeling is conducted on an impeller of a low-temperature deep-well pump; combining the impeller subjected to parametric modeling with other flow passage components, and performing numerical simulation calculation to obtain the hydraulic performance of the low-temperature deep-well pump model; the hydraulic performance comprises lift and efficiency; if the low-temperature deep-well pump model needs to be optimized, sensitivity analysis is carried out with geometric design parameters of the impeller as input parameters and hydraulic performance of the low-temperature deep-well pump model as output parameters, a multi-objective optimization MOP agent model is constructed according to a sensitivity analysis result, the multi-objective optimization MOP agent model is optimized based on a particle swarm optimization algorithm, and the low-temperature deep-well pump model is obtained. And obtaining an optimized impeller model. Compared with a traditional optimization method, the method can provide reliable design reference for designers lacking rich design experience, so that the design quality is improved, the calculation precision is ensured, the optimization process is remarkably accelerated, and the required optimization time is shortened.

The invention discloses a design method and device for an air conditioner heat exchanger of an electric vertical take-off and landing aircraft and a storage medium, and the method comprises the steps: firstly determining multiple types of geometric design parameters and parameter ranges of the heat exchanger, and building a heat exchanger weight calculation model and a heat exchanger heat exchange amount calculation model; sampling parameter ranges of various geometric design parameters to obtain N sample point sets; then, a multi-objective evaluation function of a genetic algorithm is constructed based on the heat exchanger weight calculation model and the heat exchanger heat exchange calculation model, and the genetic algorithm is adopted to carry out multi-time iterative optimization solution based on the multi-objective evaluation function by taking the N sample point sets as populations to obtain multiple groups of heat exchanger geometric design parameter optimization solution sets; and finally, verifying the thermodynamic properties under each group of heat exchanger geometric design parameter optimal solution set, so as to determine the heat exchanger geometric design parameter optimal solution set. The device and the storage medium are used for implementing the method. According to the method, the rapid iteration and design type selection requirements of the thermal management system can be met.

The embodiment of the invention provides a heat exchange equipment optimization design method and system, a medium and equipment, and the method comprises the steps that according to the design working condition of heat exchange equipment, thermodynamic parameters and geometric design parameters are determined; determining a pipe side heat transfer film coefficient and a shell side heat transfer film coefficient according to the geometric design parameters; according to the pipe side heat transfer film coefficient, the shell side heat transfer film coefficient, the heat exchanger pipeline outer side area, the heat exchanger pipeline inner side area and the pipe wall thickness, the total heat exchange coefficient is determined; according to the thermodynamic parameters and the total heat exchange coefficient, the total heat exchange area of the heat exchange equipment is determined; and the total heat exchange area is substituted into the evaluation model, and geometric design parameters corresponding to the total heat exchange area meeting preset conditions serve as target design parameters and are applied to design of the heat exchange equipment. Through the method, a heat exchange equipment design scheme considering both design flexibility and engineering feasibility can be obtained.

The invention discloses a curve adjustment method and device based on feature control points, equipment, a medium and a product, and relates to the field of computer-aided geometric design, the method comprises the following steps: introducing a shape parameter a, and constructing a planar cubic Bezier curve with the shape parameter; according to design requirements, performing interpolation processing on feature points of the planar cubic Bezier curve, and determining curve segments with the feature points; based on a continuity constraint condition, splicing all curve segments with feature points; initializing adjustment parameters according to the curve type, the feature point type and the interpolation position of the planar cubic Bezier curve; according to the initialized adjustment parameters, control points of the curve segments with the feature points are calculated, and a linear equation set is constructed; the control parameters are optimized according to the linear equation set, the curve shape is adjusted according to the optimized control parameters, curve continuity can be improved, and the type of the insertion point can be controlled.

A computer-implemented method for systematically configuring, evaluating and optimizing a heat transfer device comprising a plurality of heat transfer elements arranged in a structured manner, each heat transfer element extending along an axial path, the method comprising: starting from an initial geometric design of the plurality of heat transfer elements; a computer simulation of fluid dynamics and heat transfer at least around the heat transfer elements is performed during an intended operation of the heat transfer device using an optimization framework with a computational model to perform a geometric design optimization for each heat transfer element based on an optimization algorithm targeting a predetermined criterion, the optimization framework is configured to iteratively improve the geometric design of the respective heat transfer elements, and to allow the geometric designs of the plurality of heat transfer elements to change independently of one another during optimization.

The application discloses a multi-objective optimization design method and related device for clamping a metal wedge-shaped anchoring joint of a composite structure, and relates to the technical field of anchoring of marine engineering composite material structures. The method comprises constructing a mechanical-bonding composite three-dimensional finite element model based on the geometric components of the metal wedge-shaped anchoring joint. Numerical simulation is performed on the multiple geometric design variables through experimental design, and key design variables affecting anchoring strength, clamping stress and joint weight are screened through variance analysis. The steepest ascent method is used to determine the optimization direction, and a quantitative mapping relationship between the design variables and the optimization objectives is constructed in the optimization region using the response surface method to establish a high-precision surrogate model. The maximum anchoring strength, the minimum clamping stress and the minimum joint weight are taken as the objectives, the surrogate model is used as an objective function evaluator, a multi-objective genetic algorithm is applied for collaborative optimization, an optimal solution set is obtained, and the optimal geometric structure parameter combination of the metal wedge-shaped anchoring joint is determined according to requirements.

The present application relates to the technical field of nuclear power plant containment structure deformation measurement, and discloses a pressure deformation measurement method for a nuclear power plant containment and related devices. The method comprises obtaining the outer curved surface geometric design parameters and correction parameters of the outer dome at the outer monitoring point, and the inner curved surface geometric design parameters of the inner dome at the inner monitoring point; obtaining the strain measurement value of the outer monitoring point of the outer dome under the action of the environmental load in the state that the inner shell bears the internal pressure; determining the free displacement estimation value of the outer shell at the outer monitoring point; determining the actual displacement value of the outer shell in the environment; determining the relative displacement change between the inner monitoring point and the outer monitoring point; and determining the deformation displacement of the inner shell at the inner monitoring point under the loaded state. The present application does not need to set up components such as metal wires inside the inner shell which are easily affected by temperature, but calculates the deformation displacement of the inner shell by the strain measurement value, the relative displacement measurement value, and the geometric design parameters and correction parameters, so as to weaken the interference of temperature on the measurement result.

The application provides a building structure material quantity prediction method and device based on a knowledge embedding neural network, electronic equipment and a storage medium, wherein the method comprises the following steps: obtaining design conditions of a target building structure, the design conditions comprising seismic design conditions, wind resistance design conditions and overall geometric design conditions of the structure; obtaining a structure design scheme of the target building structure, the structure design scheme comprising the spatial positions and geometric sizes of walls, columns, beams, plates and supports; inputting the design conditions and the structure design scheme into a pre-trained material quantity prediction model to obtain the material quantities of various materials in the target building structure. The method embeds prior knowledge into the material quantity prediction model, overcomes the defects of the prior art that the material quantity prediction value violates the basic common sense of engineering and has low prediction accuracy due to the inability to consider the prior knowledge in the field related to building structures, and realizes accurate and efficient prediction of the material quantity of the target building structure.

This invention relates to the field of earthwork trenching technology, specifically to a method and system for excavating slope trenches for prefabricated quick-assembly grid beams. In this invention, by collecting high-density discrete coordinates of the slope and constructing a reference straight-line vector connecting the beginning and end, the geometric correction depth of each surface point relative to the ideal flat bottom surface is calculated. This eliminates nonlinear errors caused by the original slope undulations, ensuring that the bottom surface of the trench accurately adapts to the straight-line configuration of the prefabricated components. The resistance gradient change characteristics during the excavation process are combined with a clustering algorithm to dynamically lock the interface between soft and hard geological layers. The excavation depth required by the geometric design is fused and compared with the measured depth of the geotechnical boundary in the same coordinate system. Based on the comparison results, differentiated cutting or stripping commands are intelligently generated. When encountering hard rock layers higher than the designed bottom surface, a fuzzy PID algorithm is activated to enhance hydraulic driving force and ensure cutting efficiency. In soft soil operating areas, standard pressure is maintained to reduce energy consumption.

The disclosure provides computing apparatus, computer-implemented methods and software for autonomously generating a geometric design for a mechanical part optimised to perform an intended function by interacting with a fluid and / or particulate matter in an operating environment. The method includes iteratively generating a plurality of CAD models for the mechanical part using a CAD kernel program based on a search distribution across a parameter space of free parameters defining geometric features of the designs, generating a simulation of the performance of the modelled mechanical part for each of the plurality of generated CAD models, analysing the simulation of each CAD model for the mechanical part over time, selecting a subset of the generated CAD models having a better performance, and proceeding to the next epoch to generate, simulate and analyse evolved CAD models having candidate geometries with features based on combinations of free parameters synthesised across an updated search distribution.

The invention discloses a neural network electromagnetic performance prediction method with causality and passivity constraints, and the method comprises the steps: obtaining geometric design variables of a TSV structure to be detected, inputting the geometric design variables into a trained electromagnetic performance prediction model, and obtaining the prediction values of a plurality of scattering coefficients of the TSV structure to be detected; wherein the electromagnetic performance prediction model comprises a polar point sub-network, a residue sub-network and a prediction sub-network, and the polar point sub-network and the residue sub-network are respectively used for determining a to-be-detected polar point vector and a to-be-detected residue vector corresponding to each scattering coefficient according to geometric design variables of a to-be-detected TSV structure; the prediction sub-network is used for determining a prediction value of a scattering coefficient of a corresponding type according to a to-be-measured pole vector and a to-be-measured residue vector; the electromagnetic performance prediction model is trained through a sample pole vector and a sample residue vector, and the sample pole vector and the sample residue vector are obtained by processing a sample scattering coefficient based on a vector fitting technology. The method is low in implementation cost and good in adaptability.

The present invention provides an optimized design method for the blades of a multi-rotor bomb-dropping fire-fighting UAV, which relates to the technical field of UAVs. The present invention obtains the geometric design data of the fuselage and the blade design data, constructs a fire-fighting UAV model using CFD software, performs aerodynamic analysis using this model to obtain flight force data; generates a blade flight evaluation coefficient based on the blade material property data and the flight force data; obtains the blade flight evaluation coefficients at each rotational speed gear and analyzes them to obtain a flight quality evaluation coefficient; constructs and trains a flight quality evaluation model with the historical blade design data as the input and the corresponding historical flight quality evaluation coefficient as the label; uses the blade design data as the initial solution and maximizes the blade flight evaluation coefficient as the optimization objective, and uses the simulated annealing algorithm to optimize the blade design data to obtain the optimal blade design data.

The application relates to a spring stiffness characteristic simulation optimization method and system. The spring stiffness characteristic simulation optimization method comprises the following steps: obtaining first stiffness characteristic curve data of a spring through multi-physical field simulation, and extracting first physical form features of the spring based on the first stiffness characteristic curve data; a prediction model between spring input parameters and the first physical form features is established; in the robust design optimization iteration, corresponding second physical form features are obtained according to the geometric design parameters and the uncertainty parameters and based on the prediction model, and second stiffness characteristic curve data of the spring is reconstructed based on the second physical form features; the robust performance evaluation result of the design scheme under the influence of uncertainty is obtained through statistical analysis based on the second stiffness characteristic curve data, and the geometric design parameters are iteratively adjusted according to the robust performance evaluation result until a spring design scheme meeting the robustness requirement is obtained.

The present invention relates to a globally consistent geometric design method for blade-like surfaces based on PHT splines, which uses Polynomial Splines over Hierarchical T-meshes (PHT-splines) surfaces to fit complex blade-like surfaces and their geometric features with high precision. The main technical solution includes: through a boundary-consistent reparameterization process, the splicing and cutting parameter domain is a globally consistent complete single parameter domain, which serves as the definition domain of the PHT spline; comprehensively considering the geometric features and boundary information, rationally sampling to obtain the set of data points to be fitted and calculating the corresponding high-order geometric information; by controlling the hierarchical subdivision of the grid, blade-like surfaces represented by PHT splines are obtained with high precision fitting. The present invention solves the discontinuity problem caused by cutting and splicing in traditional design methods, ensures the global consistency of the overall design, and improves the design accuracy and efficiency.

This invention belongs to the field of road traffic noise control and acoustic computation technology, and relates to a physical information neural network-driven method for sound barrier shape optimization based on the isogeometric singular boundary method. The method includes: NURBS parameterization and boundary discretization based on the sound barrier's geometric design requirements; constructing an acoustic analysis model based on acoustic control equations and boundary conditions; calculating the average sound pressure level within the observation area using the isogeometric singular boundary method and using it as the objective function; constructing a physical information neural network, constraining the control point update range through a coordinate mapping layer and bounded activation functions, and iteratively optimizing using the average sound pressure level of the observation area as the loss function; and outputting the optimized sound barrier geometry after network convergence and evaluating the noise reduction effect. This invention combines the isogeometric singular boundary method with a physical information neural network, eliminating the need for mesh reconstruction and sensitivity analysis, achieving efficient adaptive optimization of the sound barrier geometry, and effectively improving the noise reduction performance of the sound barrier.

The present invention relates to a solution for a floating wind platform made of reinforced concrete for mass production, characterized by a geometric design providing a hydrostatic natural prestressing to the concrete, causing it to work under compression. The structural response of the platform for working in the most effective mode is improved, and the occurrence of fractures or cracks in the concrete is prevented, which reduces permeability and allows for reducing the rebar to be contained in the structure, also increasing operational safety. Furthermore, the invention has a system for anchoring the mooring lines to the structure in the form of a truss made of reinforced concrete which evenly distributes mooring stresses, minimizing prestressing in the high area of the platform, and increasing the area for distributing shear forces due to the change in section between the platform and the tower of the wind turbine. The geometric design furthermore confers the versatility of being able to adopt low draft SPAR, semi-submersible, barge, or buoy solutions, with the wind turbine being installed such that it is centered or off-center on the structure, thereby being adapted to different draft requirements or environmental and logistics conditions.

The application discloses a complex part contour optimization design method based on multi-field node feature weighted aggregation. First, according to the complex part contour geometry design variables, temperature and contact surface friction coefficient to be optimized, a batch of finite element simulations are carried out to obtain a training data set. Then, a deep map sampling aggregation neural network fusing space and temperature attention is constructed, and then the deep map sampling aggregation neural network fusing space and temperature attention is trained by using the training data set. After the training is completed, a stress prediction model is obtained. Finally, combined with the stress prediction model, an iterative optimization is carried out to minimize the maximum equivalent stress of the complex part as an optimization objective, and the optimal geometric parameters of the current complex part contour in the target temperature interval are obtained. The application realizes the rapid reconstruction of the stress distribution state of the complex part, improves the optimization speed while fully guaranteeing the accuracy of the optimization result, and greatly saves the calculation resources.