Object change prediction method, load information calculation method, program, recording medium, apparatus, material processing method, material processing apparatus, and raw material manufacturing method

Abstract

To provide a method capable of predicting a change of an object when an action is applied to the object in a relatively short time and with high accuracy.SOLUTION: A prediction method of the present invention is a method of predicting a change of an object when an action is applied to the object, and includes predicting a change of a first object in a region to which the action of the object is applied and a vicinity thereof by a graph neural network, and predicting a change of a second object in at least a part of a region other than the region to which the action of the object is applied and the vicinity thereof by analysis using a finite element method.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a method for predicting changes in an object, a method for calculating load information, a program, a recording medium, an apparatus, a material processing method, a material processing apparatus, and a manufacturing method of a raw material. [Background technology]

[0002] Various shapes of preformed materials are used in structures. These preformed materials are produced by deforming and connecting materials such as metals and resins. The deformation and connection of the above materials are carried out, for example, by applying external forces such as heating or pressing to a desired region of the material to deform or weld it in order to obtain a preformed material of the desired shape.

[0003] However, when deforming a material by heating, for example, it is necessary to set various conditions to obtain a preformed material with the desired shape, such as the location to be heated, the shape, temperature, and heating time. Furthermore, it is often difficult to obtain a preformed material with the desired shape by a single heating, and it is necessary to set the number of heatings and the heating conditions for each of the multiple heatings. Also, when connecting metal materials by welding, for example, the welding may cause distortion in the metal material, and work to remove this distortion may be required. For this reason, it is very important to predict the changes in an object, such as a material, when some kind of action is applied to the object.

[0004] As a method for predicting the physical properties of a material when a load is applied, for example, a load estimation method is known that uses a computer simulation based on the finite element method to estimate the amount of load applied to a given position on a component when an impact force is applied to the component (see Patent Document 1). [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Laid-Open No. 2013-057691 Summary of the Invention [Problem to be solved by the invention]

[0006] However, in methods for estimating changes in an object, finite element analysis, for example, methods using elasticity analysis, can predict changes in an object with high accuracy, but have the problem that it takes a long time to obtain the predicted results.

[0007] Therefore, an object of the present invention is to provide a method that can predict, in a relatively short time and with high accuracy, changes in an object when an action is applied to the object. [Means for solving the problem]

[0008] As a result of extensive research into achieving the above object, the inventors have discovered that by combining a graph neural network with analysis using the finite element method, it is possible to predict changes in an object when an action is applied to the object in a relatively short time with high accuracy. The present invention was completed based on these findings.

[0009] That is, the present invention provides a method for predicting a change in a subject when an action is applied to the subject, comprising: The present invention provides a prediction method including predicting, by a graph neural network, changes in a first object in a region of the object where the action is applied and in its vicinity, and predicting, by analysis using a finite element method, changes in a second object in at least a portion of a region other than the region of the object where the action is applied and in its vicinity.

[0010] The prediction method is a method for predicting a change in a characteristic and / or a state of the target object when an action is applied to the target object, It is preferable that the method includes predicting a change in a first property and / or state in the area of ​​the object where the action is applied and its vicinity using a graph neural network, and predicting a change in a second property and / or state in at least a part of the area of ​​the object other than the area where the action is applied and its vicinity using analysis by the finite element method.

[0011] The prediction method is a method for predicting a change in a physical property of a material when a load is applied to the material, the method comprising: The method may include predicting a change in a first physical property near a load location using a graph neural network, and predicting a change in a second physical property in at least a portion of an area other than the load location using elastic analysis using a finite element method.

[0012] the change in the first physical property includes one or more selected from the group consisting of a stress increment, a temperature increment, a stress increment, a displacement increment, an elastic strain increment, a plastic strain increment, a residual stress change, a stiffness change, and a tensile strength change; The change in the second physical property preferably includes one or more selected from the group consisting of a plastic strain increment, a stress increment, a temperature increment, a displacement increment, and a residual stress increment.

[0013] The prediction method is preferably a method for predicting changes in physical properties of the material when the material is subjected to heat processing, cooling processing, or mechanical processing due to the load.

[0014] The prediction method is preferably a method for predicting changes in the physical properties of the material when the material is subjected to one or more treatments selected from the group consisting of welding, deformation, cutting, laminating, cutting, grinding, punching, forming, hot pressing, cold pressing, and peening due to the load.

[0015] The prediction method is a method for predicting a change in a physical property of a material when a thermal load is applied to the material, the method comprising: The change in the first physical property may include a first plastic strain increment and a first temperature increment, and the change in the second physical property may include a second plastic strain increment predicted based on the first temperature increment.

[0016] The prediction method is preferably a method for predicting changes in the physical properties of the first metal material and / or the second metal material when the first metal material is welded to the second metal material under thermal load.

[0017] The prediction method is preferably a method for predicting changes in physical properties of the material when the material is deformed by a thermal load and / or a physical load.

[0018] It is preferable that the graph neural network has learned the relationship between a load applied to a local region of the material and a change in physical properties in the local region when the load is applied.

[0019] The present invention also provides a method for calculating load information, which calculates load information for obtaining a target raw material using the change in the first physical property and the change in the second physical property obtained by the prediction method.

[0020] The present invention also provides a program for executing the above prediction method.

[0021] The present invention also provides a computer-readable recording medium that stores the above program.

[0022] The present invention also provides an apparatus including a calculation unit that acquires load information using the above calculation method.

[0023] The present invention also provides a method for processing material based on load information calculated by the above calculation method.

[0024] The present invention also provides a material processing apparatus that includes a program for executing the above method.

[0025] The present invention also provides a method for manufacturing a preform, which comprises processing a material by applying a load to the material based on the load information calculated by the above-described calculation method. [Effects of the Invention]

[0026] The prediction method of the present invention makes it possible to predict, in a relatively short time and with high accuracy, changes in an object when an action is applied to the object. As a result, the prediction method of the present invention can predict changes in the physical properties of a material when a load is applied to the material, such as changes in the physical properties of a metal material when metal materials are joined by welding, or changes in the physical properties of a material when a material is deformed by being pressed with a press. [Brief explanation of the drawings]

[0027] [Figure 1] 1 is a flowchart illustrating an embodiment of the prediction method of the present invention. [Figure 2] 10A and 10B are images and graphs showing a portion of the results of transient analysis of displacement in the example. [Figure 3] 10 is an image showing a portion of an excerpt from the results of transient stress analysis in an example. [Figure 4] 10 is an image showing another portion of the results of transient stress analysis in the example. [Figure 5] 10 is an image showing another portion of the results of transient stress analysis in the example. DETAILED DESCRIPTION OF THE INVENTION

[0028] [Method for predicting changes in the target] The prediction method of the present invention is a method for predicting a change in an object when an action is applied to the object, which includes at least predicting a change in a first object in a region of the object where the action is applied and its vicinity using a graph neural network (GNN), and predicting a change in a second object in at least a part of a region other than the region where the action is applied and its vicinity using an analysis by the finite element method (FEM).

[0029] Examples of the target include objects such as materials; space; fields (target phenomena) such as deformation fields, stress fields, temperature fields, flow fields, sound fields, electric fields, magnetic fields, light fields, gravitational fields, acceleration fields, potential fields, and energy fields; and physical quantities thereof. The term "field" above can also be replaced with "distribution" or "time history." The target material of the prediction method is not particularly limited, and examples include solids such as resin, metal, wood, ceramics, composite materials, fiber materials, rubber materials, glass, and concrete. The shape of the material is also not particularly limited, and may be any of plate, powder, wire, block, film, fiber, and the like. The state of the material is not particularly limited, and may be any of solid, liquid, gas, critical fluid, and the like.

[0030] The above-mentioned actions include loads such as heat (thermal load) and external force (physical load). The actions applied to the object include both actions that are actively applied to the object and actions that are passively applied to the object. Specific examples of events that apply (or are subjected to) the above-mentioned actions include various thermal and cooling processes and mechanical processes such as welding, deformation, cutting, lamination, cutting, grinding, punching, molding, hot pressing, cold pressing, and peening; methods of applying voltage from an external source; phenomena in which thermal loads or physical loads such as heat or cooling are applied; and phenomena in which waves are transmitted through fluids, such as the transmission of sound or light through air.

[0031] The welding is the welding of two metal materials (specifically, welding a first metal material to a second metal material), and methods include multi-layer welding, friction stir welding, build-up welding, MIG welding, MAG welding, submerged arc welding, manual stick welding, laser welding, electron beam welding, and TIG welding. Examples of the deformation include pressing with a press or the like, bending, drawing, and drawing. Examples of the lamination include sheet lamination, lamination (resin film lamination and metal additive manufacturing), additive manufacturing using a 3D printer, powder additive manufacturing, and adhesive additive manufacturing.

[0032] That is, the prediction method of the present invention can predict thermoelastic and thermo-elasto-plastic phenomena that occur when active processes such as the above-mentioned thermal processing, cooling processing, and mechanical processing are performed; phenomena in which sound and light are transmitted through the air (for example, acoustic effects (sound fields) in the presence of meteorological influences (flow fields)); carrier movement (drift, diffusion) that is the rate-limiting factor for the operation of electronic devices when voltage is applied from the outside, as well as carrier distribution and fixed charge distribution due to carrier movement, and potential (electric potential) distribution based on these.

[0033] The events in the change of the object include various properties and / or states. The properties are preferably physical properties. Examples of the physical properties include, but are not limited to, temperature, stress, displacement, elastic strain, plastic strain, residual stress, stiffness, tensile strength, light, heat, sound, electric charge, carriers, voltage, current, pressure, viscosity, electricity, magnetism, and airflow. The changes in the object include the amount of change in the properties or states, such as temperature increment, stress increment, displacement increment, elastic strain increment, plastic strain increment, residual stress change, stiffness change, and tensile strength change, as well as the propagation velocity, propagation direction, and distributions including one or more of these. Note that, in this specification, the term "increment" does not only refer to the amount of increase, but also includes the amount of decrease in the case of a decrease, i.e., the "amount of change." The change in the object obtained by the prediction method of the present invention may be a change in the object from an object before an action is applied (e.g., a raw material) (initial state) to an object after an action is applied (e.g., a preformed material) (final state), or it may be a distribution of the object over part or all of the period from the initial state to the final state (transient analysis result).

[0034] One embodiment of the prediction method of the present invention will be described using the flowchart shown in Fig. 1. Fig. 1 is a flowchart for predicting changes in physical properties when a load is applied to a material. Note that the configurations shown in the drawings and in the following description are examples, and the scope of the present invention is not limited to those shown in the drawings and in the following description.

[0035] (Prediction of changes in the first physical property) In this embodiment, first, a graph neural network (GNN) (S1) predicts a change in a first physical property (a change in a first object) at a load location and its vicinity. To output the change in the first physical property, a load condition (action condition) is input to the GNN (S1). As a result, the GNN (S1) predicts and outputs a change in the first physical property at the load location and its vicinity (action area and its vicinity) caused by the input load condition.

[0036] The GNN (S1) is a trained GNN that has learned the relationship between action information and changes in the object based on the action information, specifically, the relationship between load information and changes in physical properties based on the load information. Furthermore, before using the GNN, an untrained GNN may be trained on the relationship between action information and changes in the object based on the action information. Training of the GNN can be performed by known or conventional methods, and is not particularly limited. For example, when predicting changes in the physical properties of a material, information calculated by a thermo-elastic-plastic analysis method such as the idealized solution method (FEM) (temperature field, stress field, displacement field, elastic strain field, plastic strain field, etc.) can be used as training data, which is appropriately selected depending on the type of action and change in the object.

[0037] The load information input to the GNN (S1) includes at least information on the load location, type of load, and load amount. That is, the action information input to the GNN preferably includes at least the area on which the action is applied, the type of action, and quantitative information on the action. Examples of the type of action include heat or external force. When the action is heat, the quantitative information (load amount) of the action includes heat input density and temperature. When the action is external force, the quantitative information (load amount) of the action includes pressure and total strain. The action information may also include other information. The other information is appropriately selected depending on the type of action, and includes, for example, the shape of the action (shape of the heated area, area to which pressure is applied, etc.) and the time during which the action is applied (duration of the action, start time of the action, etc.). Specifically, in the case of a thermal load, examples of the information include the heat input amount, heating rate, heating position (start point and end point), heating direction, heating depth, heating distribution, etc.

[0038] The change in the first physical property (change in the first object) output by the GNN may be one type or two or more types, and typically two or more types. Examples of the change in the first physical property output by the GNN (S1) include a temperature field, a stress field, a displacement field, an elastic strain field, and a plastic strain field. The temperature field includes a temperature increment and its transient information (distribution). The stress field includes a stress increment and its transient information (distribution). The displacement field includes a displacement increment and its transient information (distribution). The elastic strain field includes an elastic strain increment and its transient information (distribution). The plastic strain field includes a plastic strain increment and its transient information (distribution). Similarly, the transformation strain and creep strain fields include their increment and transient information (distribution).

[0039] The change in the first physical property preferably includes one or more selected from the group consisting of a temperature increment, a stress increment, a displacement increment, an elastic strain increment, a plastic strain increment, a residual stress change, a stiffness change, and a tensile strength change, and more preferably includes one or more selected from the group consisting of a temperature increment, a stress increment, a displacement increment, and a plastic strain increment.

[0040] GNNs excel at predicting local phenomena. They are also adept at communicating between mutually isolated regions. This allows them to predict changes in the target object with high accuracy in the region of action and its vicinity, while also being faster than conventional analysis methods (such as FEM thermo-elastic-plastic analysis) that are capable of highly accurate predictions. Here, "region of action and its vicinity" refers to the region that includes the location where the action is applied, and is within the range that can be predicted by the GNN. This range can be set appropriately, taking into account the balance between prediction accuracy and speed.

[0041] (Prediction of changes in the second physical property) Next, a change in a second physical property (change in the second object) in at least a portion of the region other than the region of action and its vicinity (sometimes referred to as "other region") is predicted by an FEM elasticity prediction (S2), which is a type of analysis using the finite element method (FEM). To output the change in the second physical property (change in the second object), at least a portion of the change in the first physical property (change in the first object) output by the GNN (S1) is input to the FEM elasticity prediction (S2). The FEM elasticity prediction (S2) then predicts and outputs the change in the second physical property in the other region that will result from the change in the input first physical property.

[0042] The FEM used in the FEM elasticity prediction (S2) uses an elasticity prediction model (prediction model) that has learned the relationship between load information (action information) and changes in physical properties (changes in the object) based on the load information. The relationship between the action information and changes in the object based on the action information may be determined using a known or commonly used FEM analysis.

[0043] A change in a first physical property is input into the FEM elasticity prediction (S2). The change in the first physical property input into the FEM elasticity prediction (S2) preferably includes a plastic strain field, and preferably further includes a temperature field and / or a stress field. The change in the first physical property preferably includes a plastic strain increment, a temperature increment, and / or a stress increment. In this specification, a "plastic strain increment" or a "temperature increment" that can be obtained as a change in the first physical property may be referred to as a "first plastic strain increment" or a "first temperature increment," etc., with the word "first" added.

[0044] The change in the second physical property output by the FEM elasticity prediction (S2) may be one type or two or more types, and usually two or more types. Examples of the change in the second physical property output by the FEM elasticity prediction (S2) include a temperature field, a stress field, a plastic strain field, etc., and specifically include those exemplified and explained as the change in the first physical property.

[0045] The change in the second physical property preferably includes one or more selected from the group consisting of a plastic strain increment, a stress increment, a temperature increment, a displacement increment, and a residual stress increment, and more preferably includes a plastic strain increment and at least one selected from the group consisting of a temperature increment, a stress increment, and a displacement increment. In this specification, a "plastic strain increment" or a "temperature increment" that can be obtained as a change in the second physical property may be referred to with the word "second" added, such as a "second plastic strain increment" or a "second temperature increment."

[0046] In this way, the GNN (S1) and FEM elasticity prediction (S2) can predict the physical information of the preform, including the area near the load point and other regions. This makes it possible to predict, for example, the plastic strain, stress, temperature, shape, volume, residual stress, and other properties of the preform obtained from the load information input to the GNN.

[0047] (Boundary conditions) When performing GNN (S1) or FEM elasticity prediction (S2), an analysis means may be used in which boundary conditions such as displacement fixation, force fixation, temperature fixation, stress fixation, temperature fixation, heat fixation, and velocity fixation are input to some regions. By using an analysis means with such boundary conditions input, it is possible to obtain changes in the object when some regions of the object are placed in a constrained state. The above analysis means may be a known or general prediction system, such as FEM analysis (thermal analysis, structural analysis, dynamic analysis, nonlinear analysis). The above analysis means may also be a trained system that has learned the relationship between action information and changes in the object based on the action information.

[0048] A specific embodiment of the prediction method of the present invention will be described below.

[0049] The first embodiment is a method for predicting changes in the physical properties of a material when the material is welded by applying a thermal load to it. In this embodiment, a temperature field including the temperature conditions during welding is input to a GNN (S1), which outputs at least a plastic strain increment (first plastic strain increment) and a temperature increment (first temperature increment) near a location where a thermal load for welding is applied (heated), as a change in a first physical property. The change in the first physical property may also include other changes. Then, the change in the first physical property including the first plastic strain increment and the first temperature increment is input to an FEM elasticity prediction (S2), which outputs at least an elastic strain increment (second plastic strain increment) predicted based on the first temperature increment, as a change in a second physical property. This makes it possible to predict and output the elastic strain of a preformed material obtained by the input temperature field.

[0050] The second embodiment is a method for predicting changes in the physical properties of a material when the material is deformed by applying a thermal load and / or a physical load (external force). In this embodiment, a GNN (S1) receives input data on the temperature and plastic strain fields that occur when the material is deformed, and outputs, as changes in the first physical property, at least a plastic strain increment (first plastic strain increment), a stress increment (first stress increment), and a displacement increment (first stress increment) near the location where the thermal load (heating) and / or external force is applied. The changes in the first physical property may also include other changes. Then, the changes in the first physical property, including the first plastic strain increment, the first stress increment, and the first displacement increment, are input to an FEM elasticity prediction (S2), which outputs, as changes in the second physical property, at least an elastic strain increment (second plastic strain increment), a stress increment (second stress increment), and a displacement increment (second stress increment) predicted based on the first temperature increment. This makes it possible to predict and output the deformation of the raw material that will be obtained from the input temperature field and plastic strain field.

[0051] Conventional methods for predicting changes in the physical properties of materials include predictions using FEM thermo-elasto-plastic analysis, GNN predictions, and other general machine learning methods. However, predictions using FEM thermo-elasto-plastic analysis require long calculation times and the manual creation of complex meshes. Furthermore, predictions using GNNs have problems, such as high GPU memory costs as the number of nodes increases, difficulty in stress propagation to distant nodes, difficulty in creating multi-layered prediction models, and the inability to propagate load information such as heat sources to restraint jigs. Other general machine learning predictions also have issues, such as difficulty predicting conditions outside the data range and the need to build separate prediction models for each joint shape. In contrast, the prediction method of the present invention combines GNNs with FEM elastic analysis to predict changes in the physical properties of materials when loaded. By using GNNs to predict the local interactions between thermal diffusion and plastic deformation and calculating the global stress balance using FEM, it is possible to predict changes in the physical properties of materials when loaded relatively quickly and with high accuracy. In addition, manual meshing is not required even when analyzing welding of complex structures, general shapes can be predicted with a single trained model, GPU memory costs can be reduced, and load information can be propagated to restraining jigs by using FEM for long-distance information propagation.

[0052] [Calculation method for action information] Using the change in the first object and the change in the second object obtained by the prediction method of the present invention, action information (load information) for obtaining a desired object can be calculated. The method for calculating the action information may be referred to as the "calculation method of the present invention." Specifically, according to the prediction method of the present invention, by inputting any action information, it is possible to predict information about the object to be obtained (e.g., physical information of a preformed material). The relationship between the any action information and the information about the object to be obtained can be accumulated as training data, and a system for calculating the action information (calculation method of the present invention) can be created. According to the calculation method of the present invention, by inputting desired physical information (e.g., shape) of a preformed material, it is possible to calculate load information on a material to obtain the physical information.

[0053] Based on the load information calculated by the calculation method of the present invention, material processing can be performed, such as welding of metal materials, press deformation, and additive manufacturing using a 3D printer, thereby producing a preform. The material processing can be performed by an operator or automatically by a machine. The above processing can be performed by a single load, or by multiple loads being applied sequentially or simultaneously. From the viewpoint of forming the preform into a shape closer to the target shape, it is preferable to apply multiple loads sequentially.

[0054] An example of an apparatus that automatically performs the above-described processing is an apparatus (material processing apparatus) that is equipped with a program that executes the above-described processing. The material processing apparatus may include, for example, a load unit that applies a thermal load and / or an external force load, and a control unit that controls the processing apparatus. The control unit is configured to be able to read the load information and to control the load unit to apply a load to the material in accordance with the load information.

[0055] [Recording Media] A recording medium storing the calculation method of the present invention can be obtained by storing the calculation method of the present invention in a recording medium. The recording medium is a computer-readable recording medium that stores the program.

[0056] The recording medium is a recording medium that can provide the program to a computer and cause the computer to execute the program, and examples of the recording medium include CD-ROMs, flexible disks, hard disks, magnetic tapes, magneto-optical disks, and non-volatile memory cards.

[0057] [Device] The device of the present invention is a device (computer system) equipped with a calculation unit for executing the task of acquiring the load information by the calculation method of the present invention. The device is composed of, for example, a calculation unit, a display unit, a recording medium, a keyboard, and a pointing device.

[0058] The calculation unit is a central processing unit that controls the entire computer. The display unit displays various input conditions and analysis results for the control executed by the calculation unit. The memory unit is a recording medium that stores the analysis results derived by the calculation unit. The keyboard is used by the operator to input various input conditions, etc. The pointing device consists of a mouse, trackball, etc.

[0059] The above-described configurations and combinations thereof of the present disclosure are merely examples, and additions, omissions, substitutions, and modifications of the configurations are possible as appropriate within the scope of the present disclosure. Furthermore, the present invention is not limited to the embodiments, but is limited only by the claims. [Example]

[0060] The present invention will be described in more detail below based on examples, but the present invention is not limited to these examples.

[0061] Example 1 The changes in physical properties of a 400mm x 400mm x 16mm metal plate when heated to a maximum temperature of 1500°C at the center of one side were predicted using conventional thermo-elastic-plastic FEM analysis or the prediction method of the present invention. Transient analysis results for plastic strain, displacement, and stress, as well as the final predicted shape, were obtained. While the conventional thermo-elastic-plastic FEM analysis took approximately 30 minutes, the prediction method of the present invention achieved a similar level of calculation accuracy while requiring approximately 2 seconds for the GNN prediction portion and approximately 1 minute for the FEM elastic prediction portion. Some images and graphs selected from the transient analysis results of the displacement of the metal plate are shown in Figure 2, and some images and graphs selected from the transient analysis results of the stress of the metal plate are shown in Figures 3-5. In the images of Figures 2-5, "TEP" refers only to the conventional thermo-elastic-plastic FEM analysis, while "GNN" refers to the prediction method of the present invention. [Explanation of symbols]

[0062] S1 GNN S2 FEM elasticity prediction

Claims

1. A method for predicting a change in a subject when an action is applied to the subject, comprising: A prediction method comprising: predicting changes in a first object in an area of ​​the object where the action is applied and its vicinity using a graph neural network; and predicting changes in a second object in at least a portion of an area other than the area of ​​the object where the action is applied and its vicinity using analysis by the finite element method.

2. A method for predicting changes in properties and / or states of an object of interest when an action is applied to the object, comprising:

2. The prediction method according to claim 1, comprising: predicting a change in a first property and / or state in a region of the object where the action is applied and its vicinity using a graph neural network; and predicting a change in a second property and / or state in at least a portion of a region of the object other than the region where the action is applied and its vicinity using analysis by a finite element method.

3. 1. A method for predicting changes in physical properties of a material when the material is subjected to a load, comprising:

2. The prediction method according to claim 1, comprising predicting a change in a first physical property at a load location and its vicinity using a graph neural network, and predicting a change in a second physical property in at least a portion of an area other than the load location and its vicinity using elastic analysis using a finite element method.

4. the change in the first physical property includes one or more selected from the group consisting of a stress increment, a temperature increment, a stress increment, a displacement increment, an elastic strain increment, a plastic strain increment, a residual stress change, a stiffness change, and a tensile strength change; The prediction method of claim 3 , wherein the change in the second physical property includes one or more selected from the group consisting of a plastic strain increment, a stress increment, a temperature increment, a displacement increment, and a residual stress increment.

5. The prediction method according to claim 3, which is a method for predicting changes in physical properties of the material when the material is subjected to heat processing, cooling processing, or mechanical processing due to the load.

6. 4. The prediction method according to claim 3, which is a method for predicting changes in physical properties of the material when the material is subjected to one or more processes selected from the group consisting of welding, deformation, cutting, laminating, cutting, grinding, punching, forming, hot pressing, cold pressing, and peening due to the load.

7. 1. A method for predicting changes in physical properties of a material when the material is subjected to a thermal load, comprising:

5. The prediction method of claim 4, wherein the change in the first physical property comprises a first plastic strain increment and a first temperature increment, and the change in the second physical property comprises a second plastic strain increment predicted based on the first temperature increment.

8. The prediction method according to claim 6, which is a method for predicting changes in physical properties of a first metallic material and / or a second metallic material when the first metallic material is welded to a second metallic material under a thermal load.

9. The prediction method according to claim 3, which is a method for predicting changes in physical properties of the material when the material is deformed by thermal and / or physical loads.

10. The prediction method according to claim 3 , wherein the graph neural network has learned a relationship between a load applied to a local region of a material and a change in physical properties in the local region when the load is applied.

11. A method for calculating load information, comprising calculating load information for obtaining a target preform, using the change in the first physical property and the change in the second physical property obtained by the prediction method according to any one of claims 3 to 10.

12. A program for executing the prediction method according to any one of claims 1 to 10.

13. A computer-readable recording medium storing the program according to claim 12.

14. An apparatus comprising a calculation unit that executes the calculation method according to claim 11 to obtain load information.

15. A method for processing material based on load information calculated by the calculation method according to claim 11.

16. A material processing apparatus having a program for executing the method according to claim 15 installed therein.

17. A method for manufacturing a preform, comprising processing a material by applying a load to the material based on load information calculated by the calculation method according to claim 11.

Images