System and method for structural design of large-scale radar system based on sub-modeling technique

Abstract

A structural design method for a large radar system based on a sub-modeling technique according to the present invention comprises: a step of classifying a large radar into a plurality of sub-models based on a metal processing method and a manufacturing method; a step of constructing an artificial neural network for the plurality of sub-models, taking material specifications as input, and extracting predicted values ​​of displacement and stress; a step of constructing a design database of the finite element analysis results of the large radar and training the artificial neural network; a step of calculating the residual between the learned finite element analysis results and the predicted values, and optimizing the sub-models to minimize the residual; and a step of assembling each of the optimized sub-models to complete the design of the entire large radar structure.

Description

Technology Field

[0001] The present invention relates to a structural design system and method for a large radar system based on a submodeling technique, and more specifically, to a structural design system and method for a large radar system based on a submodeling technique that can reduce the time required for design and analysis and improve structural stability by utilizing submodeling-based finite element analysis in the structural design of a large radar. Background Technology

[0003] Recently, neural network-based techniques such as PINN (Physics-Informed Neural Network) and GNN (Graph Neural Network) have been utilized in academia to solve inverse problems of physical phenomena or to efficiently analyze nonlinear problems. In this context, PINN is a technique that directly incorporates physical equations (e.g., partial differential equations) into the neural network learning process, enabling predictions that adhere to physical laws as well as existing data. GNN, on the other hand, is a technique that effectively models the interconnected behaviors or interactions of structures by learning the relationships between nodes and edges in data with complex graph structures.

[0004] These techniques are sometimes used in combination with sub-modeling methods when high-order derivative operations or large-scale computations are required. Sub-modeling is a method that divides the entire structure into detailed unit models to distribute the computational burden and reduce computation time through parallel or sequential processing.

[0005] However, conventional technologies have several limitations. First, in actual industrial environments, numerical analysis using PINN, GNN, etc., is less efficient because it often requires more computation time than conventional Finite Element Analysis (FEA). Second, accumulating training data for the entire 3D model and training it anew based on it demands very high performance from computing devices such as GPUs, making practical application difficult.

[0006] Therefore, there is a problem in that it is difficult to rapidly and efficiently optimize the design of large radar structures using only existing technology. The problem to be solved

[0008] Accordingly, the present invention aims to solve the aforementioned problems, and the objective of the present invention is to provide a structural design system and method for a large radar system based on a submodeling technique that can reduce the time required for design and analysis and improve structural stability by utilizing submodeling-based finite element analysis in the structural design of a large radar.

[0009] However, the problems that the present invention aims to solve are not limited to those described above, and other problems may exist. means of solving the problem

[0011] A method for designing the structure of a large radar system based on a sub-modeling technique according to an embodiment of the present invention for achieving the above-mentioned purpose comprises: a step of classifying a large radar into a plurality of sub-models based on a metal processing method and a manufacturing method; a step of constructing an artificial neural network for the plurality of sub-models, taking material specifications as input, and extracting predicted values ​​of displacement and stress; a step of constructing a design database of the finite element analysis results of the large radar and training the artificial neural network; a step of calculating the residual between the learned finite element analysis results and the predicted values, and optimizing the sub-models to minimize the residual; and a step of assembling each of the optimized sub-models to complete the design of the entire large radar structure.

[0012] The step of classifying a large radar into multiple sub-models based on metal processing and manufacturing methods can involve dividing the large radar into a main frame, drive unit, doors, and other devices, and configuring sub-models by detailed structural unit.

[0013] The step of constructing an artificial neural network for the plurality of sub-models, using material specifications as input, and extracting predicted values ​​for displacement and stress may involve constructing the artificial neural network independently for each of the plurality of sub-models.

[0014] The step of constructing an artificial neural network for the above-mentioned plurality of sub-models, using material specifications as input, and extracting predicted values ​​of displacement and stress can utilize the extracted displacement and stress as key parameters for verifying structural stability.

[0015] The step of building a design database with the finite element analysis results of the previously constructed large radar and training the artificial neural network includes the design database may include the finite element analysis results of a previously designed large radar.

[0016] The step of completing the design of the entire large radar structure by assembling each of the above-mentioned optimized sub-models allows the assembled entire large radar structure to respond to new design changes by changing modification variables for each sub-model without repeatedly performing existing finite element analysis.

[0017] The number of the above sub-models may be variably changed depending on the specifications and design purpose of the large radar.

[0018] The step of constructing an artificial neural network for the above plurality of sub-models, using material specifications as input, and extracting predicted values ​​of displacement and stress can extract design elements including at least one of thermal deformation, fatigue fracture, and vibration characteristics in addition to the displacement and stress.

[0019] The above design database is not limited to metal materials and can be expanded to reflect the physical properties of new materials, including composites.

[0020] The step of constructing an artificial neural network for the plurality of sub-models, taking material specifications as input, and extracting predicted values ​​of displacement and stress can derive the predicted values ​​using an artificial neural network and a partial differential equation calculator based on the material specifications and design conditions input for each sub-model.

[0021] The step of calculating the residual between the learned finite element analysis result and the predicted value, and optimizing the sub-model to minimize the residual, can be performed by calculating the residual as a sum of values ​​obtained by multiplying the error with the measured data and the violation of physical equation constraints by weights, respectively, and iteratively learning until the residual becomes less than a preset threshold to derive the final sub-model design result.

[0022] The step of calculating the residual between the learned finite element analysis result and the predicted value, and optimizing the sub-model to minimize the residual, may repeat the process of recalculating the predicted value and recalculating the residual if the calculated residual does not satisfy a preset criterion.

[0024] Meanwhile, a structural design execution system for a large radar system based on a submodeling technique according to another embodiment of the present invention comprises: a submodel classification unit that classifies the large radar into a plurality of submodels based on a metal processing method and a manufacturing method; a neural network computation unit that receives material specifications for each of the plurality of submodels as input and predicts displacement and stress; a finite element analysis learning unit that stores the results of a finite element analysis (FEA) of the large radar that has been constructed and trains the neural network computation unit; a submodel optimization unit that calculates the residual between the predicted value of the neural network computation unit and the learned finite element analysis result and optimizes the submodel to minimize the residual; and an integrated assembly unit that assembles the optimized submodels to complete the design of the entire large radar structure.

[0025] The above sub-model classification unit can classify the large radar into a main frame, drive unit, doors, and other devices, and configure sub-models by detailed structural unit.

[0026] The above artificial neural network operation unit can calculate a predicted value by independently configuring an artificial neural network for each of the multiple sub-models.

[0027] The above finite element analysis learning unit may include the finite element analysis results of a large radar designed in the past.

[0028] In some embodiments of the present invention, the integrated assembly can respond to new design changes by changing modification variables for each sub-model without repeatedly performing existing finite element analysis.

[0029] The number of the above sub-models may be variably changed depending on the specifications and design purpose of the large radar.

[0030] The above artificial neural network computation unit can extract design elements including at least one of thermal deformation, fatigue failure, and vibration characteristics in addition to displacement and stress.

[0031] The above design database is not limited to metal materials and can be expanded to reflect the physical properties of new materials, including composites.

[0032] The above artificial neural network computation unit can derive predicted values ​​using an artificial neural network and a partial differential equation (PDE) calculator based on the material specifications and design conditions input for each sub-model.

[0033] The above submodel optimization unit calculates the residual between the predicted value and the learned finite element analysis result, calculates the residual by summing the error with the measured data and the PDE constraint violation by multiplying them by weights respectively, and can derive the final submodel design result by iteratively learning until the residual becomes less than a preset threshold.

[0034] The above submodel optimization unit can repeat the process of recalculating the predicted value and recalculating the residual if the set criteria are not satisfied after calculating the residual, until each submodel design is stabilized.

[0035] A computer program according to another aspect of the present invention for solving the above-mentioned problem is combined with a computer, which is hardware, to execute a system and method for designing the structure of a large radar system based on the submodeling technique, and is stored on a computer-readable recording medium.

[0036] Other specific details of the present invention are included in the detailed description and drawings. Effects of the invention

[0038] According to the present invention, an artificial neural network is constructed for each submodel, and displacement (u) and stress are extracted using key parameters, such as material properties and specifications (size, thickness, etc.), as input values. The extracted results are trained in conjunction with the company's existing design database. Through this process, residuals are calculated between previously verified design values ​​and neural network prediction values, and optimization is performed by minimizing these residuals. The optimization method can be flexibly adjusted according to the engineer's choice, enabling customized structural optimization based on design conditions or goals. This allows for the simultaneous securing of both the performance and structural stability of each submodel.

[0039] In addition, according to the present invention, each optimized submodel is assembled to complete the entire large radar structure. This solves the problem of having to repeatedly perform the entire finite element analysis (FEA) when components are changed, which occurred in conventional design methods. By applying the present invention, parts of the four major submodels constituting the large radar that have little variation from an engineering perspective are maintained, and even when variation is required, only variables such as the size and thickness of the submodel need to be adjusted, thereby reducing the hassle and time required to manually generate a new mesh.

[0040] Furthermore, utilizing existing design databases can enhance prediction accuracy, and assembling the entire structure based on submodels with verified stability can effectively reduce the time and analysis costs required throughout the design process. This improves the structural design efficiency of large radar systems and minimizes the cost burden resulting from iterative design and analysis.

[0041] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below. Brief explanation of the drawing

[0043] The drawings attached below are intended to aid in understanding the embodiments thereof and provide embodiments together with a detailed description. However, the technical features of the embodiments thereof are not limited to specific drawings, and the features disclosed in each drawing may be combined with one another to form new embodiments. FIG. 1 is a block diagram of a structural design system for a large radar system based on a sub-modeling technique according to one embodiment of the present invention. FIG. 2 is a flowchart of a large radar structure design method according to one embodiment of the present invention. FIG. 3 is a diagram illustrating an example of an artificial neural network applied to one embodiment of the present invention. FIG. 4 is a diagram illustrating an example of applying variables to an artificial neural network in one embodiment of the present invention. FIG. 5 is a diagram illustrating the sub-modeling optimization process in one embodiment of the present invention. FIG. 6 is a diagram illustrating the residual calculation process in one embodiment of the present invention. FIG. 7 is a diagram illustrating the submodel assembly process in one embodiment of the present invention. Specific details for implementing the invention

[0044] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the present invention, and the present invention is defined only by the scope of the claims.

[0045] The terms used in this specification are for describing embodiments and are not intended to limit the invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. The terms "comprises" and / or "comprising" used in this specification do not exclude the presence or addition of one or more other components in addition to the components mentioned. Throughout the specification, the same reference numerals refer to the same components, and "and / or" includes each of the mentioned components and all combinations of one or more. Although terms such as "first," "second," etc., are used to describe various components, these components are not limited by these terms. These terms are used merely to distinguish one component from another. Therefore, the first component mentioned below may be the second component within the technical scope of the invention.

[0046] Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

[0047] An embodiment of the present invention described below can efficiently solve the problem of incurring significant design costs for structural design and finite element analysis (FEA) due to the characteristics of large radars. It reduces the problem of consuming large computer resources caused by the need to repeatedly verify results during the design and analysis processes, as well as the increase in the number of elements and complex boundary conditions resulting from the large scale.

[0048] In particular, the present invention classifies each major design part into sub-models during the structural design of a large radar, and verifies structural stability through experimental and analytical values ​​by identifying design parameters and performing finite element analysis for each sub-model. Subsequently, by assembling each sub-model to complete the design of the entire large radar structure, the invention effectively reduces the time and cost required in the conventional design and analysis process and improves structural stability.

[0049] Hereinafter, a structural design system and method for a large radar system based on a submodeling technique according to the present invention will be described in detail with reference to the attached drawings.

[0050] FIG. 1 is a block diagram of a structural design system (100) of a large radar system based on a sub-modeling technique according to one embodiment of the present invention.

[0051] Referring to FIG. 1, the large radar structure design system (100) according to the present invention includes a submodel classification unit (110), an artificial neural network operation unit (120), a finite element analysis learning unit (130), a submodel optimization unit (140), and an integrated assembly unit (150).

[0052] First, the sub-model classification unit classifies the large radar into multiple sub-models based on metal processing and manufacturing methods. This allows the entire structure to be efficiently divided, enabling design and analysis to be performed at the individual sub-model level.

[0053] The artificial neural network computation unit (120) receives material specifications for each submodel as input and predicts displacement and stress. The input values ​​include physical parameters necessary for design, such as structural characteristics, boundary conditions, and load conditions, and the structural response of the submodel can be calculated quickly and accurately through the trained neural network.

[0054] The finite element analysis learning unit (130) stores the finite element analysis (FEA) results of the previously constructed large radar and trains the artificial neural network computation unit (120) with the corresponding data. This increases the accuracy of the predicted values ​​for each sub-model and provides a learning foundation necessary for the residual calculation and optimization process.

[0055] The submodel optimization unit (140) calculates the residual between the predicted value derived from the artificial neural network operation unit (120) and the learned finite element analysis result, and iteratively optimizes the submodel to minimize the residual. The submodel optimization unit (140) can design the submodel through iterative learning based on a loss function and verification of PDE constraints.

[0056] The integrated assembly unit (150) designs the entire large radar structure by assembling optimized sub-models. Through this, the design results for each sub-model are integrated, the stability of the entire structure is ensured, and the design change can be responded to quickly without repeating the existing FEA.

[0057] Hereinafter, with reference to FIGS. 2 to 7, the operation of a large radar structure design system (100) according to one embodiment of the present invention and a structure design method of a large radar system based on a submodeling technique according to the present invention corresponding to the operation of the large radar structure design system (100) will be specifically described.

[0058] FIG. 2 is a flowchart of a large radar structure design method according to one embodiment of the present invention.

[0059] Referring to FIG. 2, first, large radars are classified into multiple sub-models based on metal processing and manufacturing methods (S110).

[0060] In step S110, as an example, the large radar can be divided into detailed structural units to define sub-models composed of a main frame, a drive unit, doors, and other devices. Specifically, the sub-models of the present invention are divided into ① a large radar main frame, ② a large radar drive unit, ③ doors, and ④ other devices. Each sub-model is distinguished according to the metal processing method and manufacturing method. In addition, the number of said sub-models can be variably adjusted according to the specifications and design purpose of the large radar. This ensures design flexibility and efficiency.

[0061] FIG. 3 is a diagram illustrating an example of an artificial neural network applied to an embodiment of the present invention. FIG. 4 is a diagram illustrating an example of applying variables to an artificial neural network in an embodiment of the present invention.

[0062] Next, an artificial neural network is constructed for the plurality of sub-models, and the material specifications are used as input to extract predicted values ​​of displacement and stress (S120).

[0063] An artificial neural network is constructed independently for each submodel. First, an artificial neural network is built for each divided submodel to derive displacement and stress, which are key parameters of structural design.

[0064] At this time, in addition to the aforementioned displacement and stress, design elements including at least one of thermal deformation, fatigue failure, and vibration characteristics can be extracted together. The material specifications input into each submodel include size (x), thickness (t), and material properties (E: Young's Modulus), and a final predicted value is calculated using an artificial neural network and a partial differential equation calculator based on the input values ​​and design conditions. Through this, the structural response and design elements of each submodel can be precisely evaluated.

[0065] Next, the finite element analysis results of the previously constructed large radar are built into a design database and trained on the artificial neural network (S130).

[0066] To this end, first, the finite element analysis results of previously designed large radars are collected and organized into a design database. This design database includes the structural analysis results of large radars designed in the past, enabling an artificial neural network to learn the structural responses of each submodel based on this data.

[0067] Furthermore, the design database can be expanded to reflect the physical properties of new materials, including composites, rather than being limited to metallic materials. This enhances prediction accuracy for various materials and design conditions, enabling the derivation of more precise prediction values ​​during the future optimization process of submodels.

[0068] Through the above process, the artificial neural network learns the relationship between the structural response for each submodel and the design parameters.

[0069] FIG. 5 is a diagram illustrating the sub-modeling optimization process in one embodiment of the present invention. FIG. 6 is a diagram illustrating the residual calculation process in one embodiment of the present invention.

[0070] Next, residuals between the learned finite element analysis (FEA) results and the predicted values ​​extracted from the artificial neural network are calculated, and each submodel is optimized to minimize said residuals (S140).

[0071] In one embodiment, the optimization process enables rapid response to new design changes by changing modification parameters for each submodel, without repeatedly performing the existing entire finite element analysis.

[0072] When calculating residuals, the error with the measured data and the degree of violation of the Partial Differential Equation (PDE) constraints are each multiplied by a weight and summed, and iterative learning is performed until the residuals fall below a preset threshold. If the residuals do not satisfy the set criteria, the process of recalculating the predicted values ​​and recalculating the residuals is repeated to derive the final submodel design result.

[0073] At this time, one embodiment of the present invention may define a loss function as shown in Equation 1 below to calculate the difference between the predicted value and the target value.

[0074] [Equation 1]

[0075]

[0076] At this time, represents the error with the measurement data (MF), and represents the degree of violation of constraints in the physical equation (PDE, Partial Differential Equation), and the total loss is the weight class It is in the form of multiplying and summing. Accordingly, iterative learning is performed until the calculated loss satisfies the criterion condition, and when the loss converges, the final submodel design result can be produced.

[0077] Referring to Fig. 6, the residual calculation process is described as follows. First, physical parameters required for design, such as structural characteristics, boundary conditions, and load conditions of each submodel, are input into an artificial neural network (S141). The input values ​​may include variables essential for structural design, such as material specifications, size, thickness, and Young's Modulus. Additionally, the learned finite element analysis results may be provided as reference data for calculations by the artificial neural network and the PDE calculator.

[0078] Next, the artificial neural network and PDE calculator computation process is performed (S142). For each divided submodel, the neural network learns non-linear relationships based on input values ​​and derives predicted values ​​for structural responses such as displacement and stress, as well as thermal deformation, fatigue, and vibration characteristics. Simultaneously, the PDE calculator evaluates the degree of constraint violation based on physical equations and calculates residuals by combining the neural network results and the PDE calculation results.

[0079] Next, a residual evaluation step is performed (S143). In this process, the calculated residuals are evaluated by comparing them with a pre-set threshold (δ), such as the mean squared error (MSE). If the residuals are below the threshold, they are determined to be at an acceptable level and confirmed as the final submodel design result. Step S143 can determine whether the design optimization criteria converge.

[0080] Finally, the optimization iteration and output step is performed (S144). If the residual does not satisfy the set threshold, the process of recalculating the predicted value and recalculating the residual is repeated. Through this iteration process, the design variables of the submodel are adjusted and optimized until the residual falls below the threshold. Once the iterative learning is complete, the design results of each submodel are finalized, and the optimized submodels are assembled to complete the design of the entire large radar structure.

[0081] FIG. 7 is a diagram illustrating the submodel assembly process in one embodiment of the present invention.

[0082] The design of the entire large radar structure is completed by assembling each of the above-mentioned optimized sub-models (S150). In step S150, the overall structural shape is constructed by reflecting the modification variables and structural responses for each sub-model optimized in the previous step. The assembly is performed considering the interfaces and boundary conditions between each sub-model, thereby completing an overall structure capable of responding to new design changes without repeating the existing finite element analysis (FEA).

[0083] Through this, the assembly process extends into integrated design data while maintaining the design parameters and physical characteristics of the submodel, ultimately ensuring the structural stability and design efficiency of the large radar.

[0084] Meanwhile, in the above description, steps S110 to S150 may be further divided into additional steps or combined into fewer steps according to an embodiment of the present invention. Also, some steps may be omitted as necessary, and the order between steps may be changed. Furthermore, even if other omitted details are omitted, the details described in FIG. 1 and the details described in FIG. 2 to 7 may be mutually applicable.

[0085] The structural design system (100) and method of a large radar system based on a submodeling technique according to one embodiment of the present invention described above may be implemented as a program (or application) and stored on a medium to be executed in combination with a computer, which is hardware.

[0086] The aforementioned program may include code encoded in computer languages ​​such as C, C++, JAVA, Ruby, and machine language, which can be read by the computer's processor (CPU) through the computer's device interface, in order for the computer to read the program and execute the methods implemented in the program. Such code may include functional code related to functions that define the necessary functions for executing the methods, and may include control code related to execution procedures necessary for the computer's processor to execute the functions according to a predetermined procedure. Additionally, such code may further include memory reference code regarding where (address) additional information or media necessary for the computer's processor to execute the functions should be referenced in the computer's internal or external memory. In addition, if the processor of the computer needs to communicate with any other computer or server located remotely in order to execute the above functions, the code may further include communication-related code regarding how to communicate with any other computer or server located remotely using the communication module of the computer, and what information or media to transmit or receive during communication.

[0087] The above-mentioned storage medium refers to a medium that stores data semi-permanently and is readable by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, examples of the above-mentioned storage medium include, but are not limited to, ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage device. That is, the above-mentioned program may be stored on various recording media on various servers that the computer can access, or on various recording media on the user's computer. Additionally, the above-mentioned medium may be distributed across networked computer systems, and computer-readable code may be stored in a distributed manner.

[0088] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0089] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present invention. Explanation of the symbols

[0091] 100: Large Radar Structural Design System 110: Submodel Classification Section 120: Artificial neural network computation unit 130: Finite Element Analysis Learning Section 140: Submodel Optimization Section 150: Integrated Assembly Unit

Claims

Claim 1 A method for designing the structure of a large radar system based on a sub-modeling technique, comprising: a step of classifying the large radar into multiple sub-models based on metal processing and manufacturing methods in a sub-model classification unit; a step of constructing an artificial neural network for the multiple sub-models in an artificial neural network computation unit, taking material specifications as input, and extracting predicted values ​​of displacement and stress; a step of constructing the results of the finite element analysis of the large radar in a finite element analysis learning unit into a design database and training the artificial neural network in the artificial neural network; a step of calculating the residual between the learned finite element analysis results and the predicted values ​​in a sub-model optimization unit, and optimizing the sub-models to minimize the residual; and a step of assembling each of the optimized sub-models in an integrated assembly unit to complete the design of the entire large radar structure, wherein the step of assembling each of the optimized sub-models to complete the design of the entire large radar structure corresponds to a new design change by changing modification variables for each sub-model without repeating the existing finite element analysis of the assembled entire large radar structure. Claim 2 A method for designing a large radar structure according to claim 1, wherein the step of classifying a large radar into multiple sub-models based on a metal processing method and a manufacturing method is to classify the large radar into a main frame, a drive unit, doors, and other devices, and configure sub-models according to detailed structural units. Claim 3 A large radar structure design method according to claim 1, wherein the step of constructing an artificial neural network for the plurality of sub-models, using material specifications as input, and extracting predicted values ​​of displacement and stress is such that the artificial neural network is constructed independently for each of the plurality of sub-models. Claim 4 A large radar structure design method according to claim 1, wherein the step of constructing an artificial neural network for the plurality of sub-models, using material specifications as input, and extracting predicted values ​​of displacement and stress utilizes the extracted displacement and stress as key parameters for verifying structural stability. Claim 5 A method for designing a large radar structure according to claim 1, wherein the step of building a design database of the finite element analysis results of the large radar previously built and training the artificial neural network comprises the design database including the finite element analysis results of the large radar previously designed. Claim 6 delete Claim 7 A method for designing a large radar structure according to claim 1, wherein the number of sub-models can be variably changed according to the specifications and design purpose of the large radar. Claim 8 A large radar structure design method according to claim 1, wherein the step of constructing an artificial neural network for the plurality of sub-models, using material specifications as input, and extracting predicted values ​​of displacement and stress involves extracting design elements including at least one of thermal deformation, fatigue fracture, and vibration characteristics in addition to the displacement and stress. Claim 9 A large radar structure design method according to claim 1, wherein the design database is not limited to metal materials but is expanded to reflect the physical properties of new materials including composite materials. Claim 10 A method for designing a large radar structure according to claim 1, wherein the step of constructing an artificial neural network for the plurality of sub-models, taking material specifications as input, and extracting predicted values ​​of displacement and stress involves deriving the predicted values ​​using an artificial neural network and a partial differential equation calculator based on the material specifications and design conditions input for each sub-model. Claim 11 A method for designing a large radar structure according to claim 10, wherein the step of calculating a residual between the learned finite element analysis result and the predicted value and optimizing the sub-model to minimize the residual is to calculate the residual by summing the values ​​obtained by multiplying the error with the measured data and the violation of physical equation constraints by weights, and to derive a final sub-model design result by iterative learning until the residual becomes less than a preset threshold. Claim 12 A large radar structure design method according to claim 11, wherein the step of calculating the residual between the learned finite element analysis result and the predicted value and optimizing the sub-model to minimize the residual involves repeating the process of recalculating the predicted value and recalculating the residual if the calculated residual does not satisfy a preset criterion. Claim 13 A system for performing structural design of a large radar system based on a submodeling technique, comprising: a submodel classification unit that classifies the large radar into multiple submodels based on a metal processing method and a manufacturing method; a neural network computation unit that receives material specifications for each of the multiple submodels as input and predicts displacement and stress; a finite element analysis learning unit that stores the results of a finite element analysis (FEA) of the large radar that has been established in a design database and trains the neural network computation unit; a submodel optimization unit that calculates the residual between the predicted value of the neural network computation unit and the learned finite element analysis result and optimizes the submodel to minimize the residual; and an integrated assembly unit that assembles the optimized submodels to complete the design of the entire large radar structure, wherein the integrated assembly unit responds to new design changes by changing modification variables for each submodel without repeatedly performing the existing finite element analysis. Claim 14 In Clause 13, the above sub-model classification unit is a large radar structural design system that classifies the large radar into a main frame, drive unit, doors, and other devices, and configures sub-models by detailed structural unit. Claim 15 In claim 13, the artificial neural network computation unit is a large radar structure design system that independently configures an artificial neural network for each of the multiple sub-models to calculate a predicted value. Claim 16 In paragraph 13, the finite element analysis learning unit is a large radar structural design system that includes the finite element analysis results of a large radar designed in the past. Claim 17 delete Claim 18 A large radar structural design system in which, in Clause 13, the number of sub-models can be variably changed according to the specifications and design purpose of the large radar. Claim 19 In claim 13, the artificial neural network computation unit extracts design elements including at least one of thermal deformation, fatigue fracture, and vibration characteristics in addition to displacement and stress in a large radar structural design system. Claim 20 In Clause 13, the above-mentioned design database is a large radar structure design system that is not limited to metal materials but is expanded to reflect the physical properties of new materials including composite materials. Claim 21 In claim 13, the artificial neural network computation unit derives predicted values ​​using an artificial neural network and a partial differential equation (PDE) calculator based on material specifications and design conditions input for each sub-model, in a large radar structure design system. Claim 22 A large radar structure design system according to claim 21, wherein the submodel optimization unit calculates the residual between the predicted value and the learned finite element analysis result, calculates the residual by multiplying the error with the measured data and the PDE constraint violation by a weight and summing the values, and iterates learning until the residual becomes less than a preset threshold to derive the final submodel design result. Claim 23 A large radar structure design system according to claim 21, wherein the submodel optimization unit repeats the process of recalculating the predicted value and recalculating the residual if the set criteria are not satisfied after calculating the residual, until each submodel design is stabilized.

Images