Metal additive manufacturing inherent strain prediction method, system, storage medium and device

Abstract

The present application relates to the field of additive manufacturing technology, and more specifically, to a method, system, storage medium and device for predicting inherent strain in metal additive manufacturing. A method for predicting inherent strain in metal additive manufacturing, comprising the following steps: obtaining initial manufacturing parameters and selecting target manufacturing parameters therefrom, dimensionlessly transforming the corresponding initial inherent strain to obtain non-quantized inherent strain, and determining a target parameter combination. An initial prediction model is constructed with the target parameter combination as input and the non-quantized inherent strain as output, and a target prediction model is obtained after dimensionality reduction processing. The target manufacturing parameters of the inherent strain to be determined are predicted using the target prediction model to obtain the target inherent strain; and the target process parameters are adjusted according to the evaluation results of the target inherent strain. According to the prediction method, a prediction system, storage medium and device are also provided, which can provide guidance with relatively clear physical meaning compared to general data-driven methods.

Description

Technical Field

[0001] The present application relates to the field of additive manufacturing technology, and more specifically, to a method, system, storage medium and device for predicting inherent strain in metal additive manufacturing. Background Art

[0002] Additive manufacturing technology, also known as 3D printing technology, is a rapid prototyping technology that builds three-dimensional objects by adding materials layer by layer. Unlike traditional subtractive manufacturing (such as milling, turning, etc.), additive manufacturing builds objects from scratch, has a high degree of design freedom and material utilization, and reduces the constraints of traditional manufacturing technology in the optimization design of complex structures.

[0003] Currently, it has been gradually applied in the fields of aerospace, medicine, and biology. Metal additive manufacturing laser selective melting technology (Selective Laser Melting, SLM) is an advanced additive manufacturing technology that uses a high-energy laser beam to selectively melt metal powder and stack it layer by layer to form three-dimensional metal parts. In the process of metal additive manufacturing laser selective melting, residual stress and deformation are two key issues that have an important impact on the performance and quality of the final parts. Residual stress refers to the stress inside the material in the absence of external loads, while deformation refers to the changes in size and shape of the parts due to thermal stress and residual stress during the additive manufacturing process.

[0004] Understanding the changing rules of residual stress and deformation and effectively controlling the residual stress and deformation is conducive to improving product quality, so it is a key issue in laser selective melting technology. The inherent strain method is an important method for predicting residual stress and deformation of formed parts in metal welding and additive manufacturing, and obtaining inherent strain is the basis and prerequisite for using the inherent strain method to analyze residual stress and deformation.

[0005] At present, the methods for extracting inherent strain mainly include experimental measurement, integral calculation, numerical simulation inverse, and data-driven methods. Among them, the experimental measurement method requires high experimental costs and takes a long time; the integral calculation method is relatively complex and has low feasibility for additive manufacturing scenarios; the numerical simulation inverse method requires the establishment of a complex model, resulting in long calculation time and high calculation cost. Therefore, it is highly necessary to establish an efficient proxy model through a data-driven method to quickly predict the inherent strain, which is very necessary for extracting the inherent strain. This method does not require modeling of complex physical processes. It directly analyzes the inherent strain data of experimental data or numerical simulation, and can relatively quickly and accurately learn the changing law of the inherent strain, thereby predicting the prediction target under given parameters, which provides effective help for researchers to grasp the physical laws and optimize the forming quality. However, the data-driven methods in the prior art are not very explanatory and the physical meaning is relatively vague. Summary of the invention

[0006] The purpose of the present invention is to overcome the deficiencies in the prior art and to provide a method that can solve the technical problem in the prior art that the physical meanings of the inherent strain evolution law and parameter adjustment law obtained are relatively vague due to the poor interpretability of the relevant technology.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions: a method for predicting inherent strain in metal additive manufacturing, comprising the following steps: S1. Obtaining initial manufacturing parameters D1: the initial manufacturing parameters D1 include various process parameters and material property parameters; S2. Obtaining target parameter combination D3: selecting target manufacturing parameters D2 from the initial manufacturing parameters D1, and using the target manufacturing parameters D2 to dimensionlessly transform the initial inherent strain I1 to obtain the dimensionless inherent strain I2, and determining the target parameter combination D3 corresponding to the dimensionless inherent strain I2 in the initial manufacturing parameters D1; ​​S3. Obtaining a target prediction model P: using the target parameter combination D3 as input and the dimensionless inherent strain I2 as output, constructing an initial prediction model P1, and performing dimensionality reduction processing on the initial prediction model P1 to obtain a target prediction model P;

[0008] The target prediction model P is shown in formulas (1)-(3):

[0009]

[0010] In the formula, represent the target dimensionless number groups related to the inherent strains in three directions, They represent the target dimensionless number group G related to the inherent strains in three directions respectively;

[0011] yes The polynomial form of yes The polynomial form of yes The polynomial form of ;

[0012] express Regression prediction model coefficients,

[0013] express Regression prediction model coefficients,

[0014] express Regression prediction model coefficients;

[0015] The calculation formula of the determination coefficient of the prediction model P is shown in formula (4):

[0016]

[0017] In the formula, represents the mean of the true output of the model, Represents the model output variable The predicted value of represents the total residual sum of squares, represents the total sum of squares of deviations;

[0018] S4. Obtaining the target inherent strain I3: predicting the target manufacturing parameter D2 of the inherent strain to be determined using the target prediction model P to obtain the target inherent strain I3; S5. According to the evaluation result of the target inherent strain I3, completing the adjustment of the target process parameter D4 among the multiple target manufacturing parameters D2.

[0019] Preferably, the step S2 includes: the initial manufacturing parameters D1 include multiple parameters, and a parameter is selected from the multiple initial manufacturing parameters D1 as a target manufacturing parameter D2; the target manufacturing parameter D2 is used as a characteristic parameter, and the initial inherent strain I1 is dimensionless to obtain the dimensionless inherent strain I2, and the dimensionless inherent strain I2 is used to evaluate the residual stress and deformation of the formed part; the target parameter combination D3 is selected from the multiple initial manufacturing parameters D1, wherein the number of the target parameter combinations D3 is the same as the type and quantity of the dimensionless inherent strain I2, and the target parameter combination D3 corresponds to the type of the dimensionless inherent strain I2 one by one.

[0020] Preferably, in the step S3, the number of the target parameter combinations D3 is determined according to the number of the dimensionless inherent strains I2; a plurality of one-to-one corresponding target parameter combinations D3 and dimensionless inherent strains I2 are respectively used as input and output to construct a plurality of initial prediction models P1; the number N of initial dimensionless number groups corresponding to the target parameter combination D3 is determined according to the independent dimensions and independent variables corresponding to the plurality of target parameters in the target parameter combination D3; an active subspace dimensionality reduction method is used to obtain a plurality of initial dimensionless number groups G1 sorted by importance; and the target prediction model P is constructed based on the plurality of initial dimensionless number groups.

[0021] More preferably, the method for constructing the target prediction model P based on multiple initial dimensionless number groups G1 specifically includes: constructing an initial polynomial model P2; determining the multiple initial dimensionless number groups G1 as the input of the initial polynomial model P2, determining the dimensionless inherent strain I2 as the output of the initial polynomial model P2, using the polynomial ridge regression method to train the initial polynomial model P2, and using the cross-validation method with the determination coefficient as an indicator until the determination coefficient corresponding to the initial polynomial model P2 is the largest, determining the target order of the polynomial and the multiple target dimensionless number groups G retained in the multiple initial dimensionless number groups G1; determining the multiple target dimensionless number groups G as input, and determining the dimensionless inherent strain I2 as output, to construct the target prediction model P.

[0022] More preferably, the number N of initial dimensionless number groups corresponding to the target parameter combination D2 is determined in the following manner, including: determining the number N1 of independent dimensions corresponding to multiple target parameters in the target parameter combination D3, determining the number N2 of independent variables corresponding to multiple target parameters in the target parameter combination D3, and determining the number N of the multiple initial dimensionless number groups by the difference between N2 and N1.

[0023] On the other hand, the present invention also provides a metal additive manufacturing inherent strain prediction system, including an acquisition module, a dimensionless module, a construction module, a prediction module and an adjustment module, wherein the acquisition module is used to acquire an initial manufacturing parameter D1 of additive manufacturing, wherein the initial manufacturing parameter D1 includes a process parameter and a material property parameter; the dimensionless module is used to select a target manufacturing parameter D2 from the multiple initial manufacturing parameters D1, and use the target manufacturing parameter D2 to dimensionlessly transform the inherent strain I1 to obtain the dimensionless inherent strain I2, and determine a target parameter combination D3 corresponding to the dimensionless inherent strain I2 in the multiple manufacturing parameters D2; the construction module is used to construct an initial prediction model P1 with the target parameter combination D3 as input and the dimensionless inherent strain I2 as output, and perform dimensionality reduction processing on the initial prediction model P1 to obtain a target prediction model P; the prediction module is used to use the target prediction model P to transform the multiple target manufacturing parameters D2 of the inherent strain to be determined into the target prediction model P Prediction is performed to obtain the target inherent strain I3 of the molten pool to be determined; an adjustment module is used to complete the adjustment of the target process parameter D4 among the multiple target manufacturing parameters D2 according to the evaluation result of the target inherent strain I3.

[0024] Preferably, the construction module includes a construction submodule and a dimensionality reduction submodule, and the construction submodule is used to determine the number of the target parameter combinations D3 according to the number of the dimensionless inherent strains I2; when the number of the target parameter combinations D3 is multiple, the multiple target parameter combinations D3 are determined as inputs and the multiple dimensionless inherent strains I2 are outputs, and multiple initial prediction models P1 are constructed.

[0025] Preferably, the dimension reduction submodule is used to determine the number N of dimensionless number groups corresponding to the target parameter combination D3 according to the independent dimensions and independent variables corresponding to the multiple target production parameters D2 in the target parameter combination D3; use the active subspace dimensionality reduction method to obtain multiple target dimensionless number groups G sorted by importance; and construct the target prediction model P based on the multiple target dimensionless number groups G.

[0026] Preferably, the dimension reduction submodule includes a construction unit, a first determination unit and a second determination unit, wherein the construction unit is used to construct an initial polynomial model P2, determine the multiple initial dimensionless number groups G1 as the input of the initial polynomial model P2, and determine the dimensionless inherent strain I2 as the output of the initial polynomial model P2.

[0027] Preferably, the first determination unit is used to determine the multiple target dimensionless number groups G as input, the dimensionless inherent strain I2 as output, and construct the target prediction model P; the second determination unit determines the number N1 of the multiple independent dimensions; determines the number N2 of the multiple independent variables; and determines the number N of the multiple dimensionless number groups by the difference between N2 and N1.

[0028] On the other hand, the present invention also provides a non-volatile storage medium, in which a program is stored, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute a metal additive manufacturing inherent strain prediction method involved in the present invention.

[0029] On the other hand, the present invention also provides a computer device, comprising a memory and a processor, wherein the processor is used to run a program stored in the memory, wherein when the program is run, a method for predicting inherent strain in metal additive manufacturing involved in the present invention is executed.

[0030] The beneficial effects of the present invention are as follows: by obtaining the target inherent strain and the evaluation result of the target inherent strain, the target process parameters in the multiple target manufacturing parameters are adjusted, the physical meaning of the target process parameter adjustment law is clarified, and the technical problem that the physical meaning of the inherent strain evolution law and parameter adjustment law obtained is relatively vague due to the poor interpretability of the relevant technology is solved. This method combines the advantages of data-driven methods and dimensionless analysis methods. Compared with traditional dimensionless analysis methods, it determines a series of dimensionless number groups that are important for inherent strain and gives their importance ranking; compared with general data-driven methods, it can give guidance with relatively clear physical meaning. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flow chart of the prediction method of the present invention;

[0032] Figure 2 It is a schematic diagram of the structure of the prediction system in the present invention;

[0033] Figure 3 It is a hardware structure block diagram of the computer device in the present invention.

[0034] Reference numerals:

[0035] 10. Computer device; 102. Processor; 104. Memory; 106. Transmission device. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0037] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0038] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the structure of the present invention.

[0039] Embodiment 1: Figure 1 As shown, the present invention provides a method for predicting inherent strain in metal additive manufacturing, comprising the following steps: S1. Obtaining initial manufacturing parameters D1: the initial manufacturing parameters D1 include various process parameters and material property parameters; S2. Obtaining target parameter combination D3: selecting target manufacturing parameters D2 from the initial manufacturing parameters D1, and using the target manufacturing parameters D2 to dimensionlessly transform the initial inherent strain I1 to obtain the dimensionless inherent strain I2, and determining the target parameter combination D3 corresponding to the dimensionless inherent strain I2 in the initial manufacturing parameters D1; ​​S3. Obtaining a target prediction model P: using the target parameter combination D3 as input and the dimensionless inherent strain I2 as output, constructing an initial prediction model P1, and performing dimensionality reduction processing on the initial prediction model P1 to obtain a target prediction model P;

[0040] The target prediction model P is shown in formulas (1)-(3):

[0041]

[0042] In the formula, represent the target dimensionless number groups related to the inherent strains in three directions, They represent the target dimensionless number group G related to the inherent strains in three directions respectively;

[0043] yes The polynomial form of yes The polynomial form of yes The polynomial form of ;

[0044] express Regression prediction model coefficients,

[0045] express Regression prediction model coefficients,

[0046] express Regression prediction model coefficients;

[0047] In the formula, represents the mean of the true output of the model, Represents the model output variable The predicted value of represents the total residual sum of squares, Represents the total sum of squares of deviations.

[0048] S4. Obtaining the target inherent strain I3: predicting the target manufacturing parameter D2 of the inherent strain to be determined using the prediction model P to obtain the target inherent strain I3; S5. According to the evaluation result of the target inherent strain I3, completing the adjustment of the target process parameter D4 among the multiple target manufacturing parameters D2.

[0049] Through the above steps, it is possible to achieve dimensionless transformation of the initial inherent strain I1 by using a variety of initial manufacturing parameters D1 to obtain dimensionless inherent strain I2, and to construct a target prediction model P with the target parameter combination D3 corresponding to the dimensionless inherent strain I2 as input and the dimensionless inherent strain I2 as output. Thus, the target inherent strain I3 to be determined is obtained by analysis using the target prediction model P, and the target process parameter D4 is adjusted according to the evaluation result of the target inherent strain I3. The purpose of clarifying the physical meaning of the target process parameter adjustment law is achieved, and the technical problem that the physical meaning of the inherent strain evolution law and parameter adjustment law obtained is relatively vague due to the poor interpretability of the relevant technology is solved.

[0050] It should be noted that in the related technology, the parameter adjustment scheme of the laser selective melting technology for metal additive manufacturing also includes: numerical simulation method. However, the numerical simulation method can conduct a more comprehensive analysis of the production process at a lower experimental cost. However, the metal additive manufacturing process is a multi-scale, multi-physical field coupling process. In terms of scale, it is necessary to consider the microscale, mesoscale, and macroscale; in terms of physical fields, it is necessary to consider the simulation of various physical processes such as heat transfer and phase change. Therefore, if a comprehensive and reliable analysis of the manufacturing process is required, there are also technical problems of low efficiency caused by long calculation time and high calculation cost.

[0051] In step S1, a plurality of initial manufacturing parameters D1 can be obtained by collecting historical experimental data, wherein the plurality of process parameters at least include: laser power (in is the laser absorption rate), scanning speed V, spot diameter , preheating temperature T0; the material physical characteristic parameters at least include: thermodynamic parameters, hydrodynamic parameters and kinetic parameters, among which the thermodynamic parameters include: specific heat capacity C P ,density , equivalent temperature change , thermal conductivity K. is the solid-state specific heat capacity, is the solid density, is the solid thermal conductivity. The melting equivalent temperature difference is , is the liquidus temperature, is the latent heat of melting; the mechanical parameters include: thermal expansion coefficient , Young's modulus E, Poisson's ratio V, yield stress .

[0052] In step S2, the initial manufacturing parameters D1 include multiple parameters, and a parameter is selected from the multiple initial manufacturing parameters D1 as a target manufacturing parameter D2; the target manufacturing parameter D2 is used as a characteristic parameter, and the initial inherent strain I1 is dimensionless to obtain the dimensionless inherent strain I2, and the dimensionless inherent strain I2 is used to evaluate the residual stress and deformation of the formed part; the target parameter combination D3 is selected from the multiple initial manufacturing parameters D1, wherein the number of the target parameter combinations D3 is the same as the type and quantity of the dimensionless inherent strain I2, and the target parameter combination D3 corresponds to the type of the dimensionless inherent strain I2.

[0053] In step S3, the number of target parameter combinations D3 is determined according to the number of dimensionless inherent strains I2; multiple one-to-one corresponding target parameter combinations D3 and dimensionless inherent strains I2 are used as input and output respectively to construct multiple initial prediction models P1. According to the independent dimensions and independent variables corresponding to the multiple target parameters in the target parameter combination D3, the number N of initial dimensionless number groups corresponding to the target parameter combination D3 is determined; the active subspace dimensionality reduction method is used to obtain multiple initial dimensionless number groups G1 sorted by importance; the target prediction model P is constructed based on the multiple initial dimensionless number groups.

[0054] Since the initial inherent strain I1 has no dimension, there is no need to select specific characteristic parameters to make it dimensionless. The dimensionless inherent strains in the three directions are , determine the input target parameter combination D3 as . The model relationship from input to output is established as shown in formulas (5)-(7):

[0055]

[0056] The target prediction model P can be obtained by reducing the dimension of the initial prediction model P1. An optional dimensionality reduction method includes:

[0057] According to the independent dimensions and independent variables corresponding to the multiple target production parameters D2 in the target parameter combination D3, the initial dimensionless number group G1 corresponding to the target parameter combination D3 is determined; the active subspace dimensionality reduction method is used to obtain multiple initial dimensionless number groups G1 sorted by importance; and the target prediction model P is constructed based on the multiple initial dimensionless number groups G1.

[0058] The number N of the initial dimensionless number group can be determined by: determining a first number of a plurality of first independent dimensions; determining a second number of a plurality of first independent variables; and determining the number of the plurality of first dimensionless number groups by using the difference between the second number and the first number.

[0059] In actual application scenarios, for a system with n independent variables and k independent dimensions, the system can be represented by at most p=nk dimensionless number groups.

[0060] For example, in the target prediction model P, there are 4 independent dimensions and 11 independent variables. The independent dimensions include length, mass, time and temperature. The 11 independent variables corresponding to the inherent strain prediction model correspond to the 11 target parameters in the target parameter combination. Then the number of dimensionless groups N = the number of independent variables N2 - the number of independent dimensions N1 = 7.

[0061] Through the active subspace dimensionality reduction method, seven initial dimensionless number groups G1 related to the inherent strain are obtained and sorted by importance.

[0062] In this embodiment, the method for constructing the target prediction model P based on multiple initial dimensionless number groups G1 specifically includes: constructing an initial polynomial model P2; determining the multiple initial dimensionless number groups G1 as the input of the initial polynomial model P2, determining the dimensionless inherent strain I2 as the output of the initial polynomial model P2, using the polynomial ridge regression method to train the initial polynomial model P2, and using the cross-validation method with the determination coefficient as an indicator until the determination coefficient corresponding to the initial polynomial model P2 is the largest, determining the target order of the polynomial and the multiple target dimensionless number groups G retained in the multiple initial dimensionless number groups G1; determining the multiple target dimensionless number groups G as input, and determining the dimensionless inherent strain I2 as output, to construct the target prediction model P.

[0063] Among them, the coefficient of determination R 2 The calculation formula is shown in formula (4),

[0064]

[0065] In the formula, represents the mean of the true output of the model, Represents the model output variable The predicted value of represents the total residual sum of squares, represents the total sum of squares of deviations;

[0066] Table 1 shows the inherent strain target prediction model P in three directions:

[0067]

[0068] The plurality of target dimensionless number groups G are determined as input, the dimensionless inherent strain I2 is determined as output, and the target prediction model P is constructed.

[0069] Then the target prediction model P is represented by formulas (1)-(3):

[0070]

[0071] In the formula, represent the target dimensionless number groups related to the inherent strains in three directions, They represent the target dimensionless number group G related to the inherent strains in three directions respectively;

[0072] yes The polynomial form of yes The polynomial form of yes The polynomial form of ;

[0073] express Regression prediction model coefficients, express Regression prediction model coefficients, express Regression prediction model coefficients;

[0074] In step S4, the multiple target manufacturing parameters D2 of the molten pool to be determined include: specifying materials, such as SS316L, Ti4Al4V, IN718, IN625 and other common materials for metal additive manufacturing, and giving their material property parameters; giving process parameters, including: laser power , scanning speed V, spot diameter , preheating temperature Etc., by inputting it into the target prediction model P for analysis, the target inherent strain I3 of the molten pool to be determined can be obtained.

[0075] For example, if the material is set to SS316L, the laser power is 400W, the scanning speed is 200mm / s, the spot size is 80um, and the preheating temperature is 20℃, the predicted target inherent strains I3 in the three directions are -0.0793 ( ), -0.0355 ( ), -0.3310 .

[0076] Then, the process may proceed to step S5, and according to the predicted target inherent strain I3, the target process parameter D4 may be adjusted to make the template inherent strain I3 approach an ideal state.

[0077] For example, if the predicted target inherent strain I3 is too high, you can reduce the laser power, increase the scanning speed, etc. to re-predict and observe whether the target inherent strain I3 under the new target process parameter D4 is ideal; or change the material, that is, change the material's physical properties, and re-predict the melt width and depth. For example, you can adjust the scanning speed to 2000mm / s,

[0078] Re-prediction, the predicted target inherent strains I3 in the three directions are -0.0023 ( )、-0.0010( )、-0.0094 ;

[0079] The laser power can also be adjusted to 40W and the prediction can be repeated. The predicted template inherent strains I3 in the three directions are -0.0079 ( )、-0.0036( )、-0.0331 After adjusting the parameters, the predicted target inherent strain I3 result is relatively ideal.

[0080] Furthermore, the trained template prediction model P can be used to predict the target inherent strain I3 under multiple target parameter combinations D3. The process parameters corresponding to the ideal inherent strain can be used as ideal process parameters to provide a reference for experimental design. Experiments are carried out according to the adjusted process parameters, and the experimental data can be added to the database as new samples to update the training model, so as to grasp the physical laws more accurately.

[0081] It is understandable that dimensional analysis is an important means of similarity research. By giving a series of dimensionless numbers with clear physical meanings, complex physical laws can be represented in a relatively concise form, and similarities in seemingly unrelated physical phenomena can be discovered. For example, the Reynolds number, which is common in fluid mechanics, can be used to characterize the flow of fluids and distinguish between laminar flow and turbulent flow in pipe flow. However, traditional dimensional analysis also has some shortcomings. For example, using its core theory, the Buckingham-PI theorem, it can only give a series of feasible solutions that meet the dimensional constraints, and it is difficult to determine the final form of the dimensionless number group; at the same time, it cannot give the importance ranking of the dimensionless number group.

[0082] The prediction method of inherent strain in metal additive manufacturing proposed in this application establishes a complete data-driven dimensional analysis framework by introducing the Active Subspace Method. The present invention uses a data-driven dimensional analysis method to establish a model for the inherent strain of metal materials, and obtains dimensionless numbers and their scaling rates that have an important influence on the inherent strain during laser selective melting. Based on the trained model, the inherent strain under given material parameters and process parameters can be predicted, the residual stress and deformation of the formed part can be evaluated, and a reference and guidance can be provided for the adjustment of process parameters, helping to determine a better process parameter window, thereby improving the performance of the formed part and optimizing the quality of the formed part. The present invention fully integrates the advantages of data-driven methods and dimensionless analysis methods. Compared with traditional dimensionless analysis methods, it can determine a series of dimensionless number groups that are important for inherent strain and give their importance ranking; compared with general data-driven methods, it can give guidance with relatively clear physical meaning.

[0083] Embodiment 2: Figure 2 As shown, a prediction system for inherent strain in an additive manufacturing process based on the prediction method includes an acquisition module, a dimensionless module, a construction module, a prediction module and an adjustment module.

[0084] Among them, the acquisition module is used to obtain the initial manufacturing parameters D1 of additive manufacturing, and the initial manufacturing parameters D1 include process parameters and material property parameters; the dimensionless module is used to select the target manufacturing parameter D2 from the multiple initial manufacturing parameters D1, and use the target manufacturing parameter D2 to dimensionlessly transform the inherent strain I1 to obtain the dimensionless inherent strain I2, and determine the target parameter combination D3 corresponding to the dimensionless inherent strain I2 in the multiple manufacturing parameters D2; the construction module is used to construct the initial prediction model P1 with the target parameter combination D3 as input and the dimensionless inherent strain I2 as output, and perform dimensionality reduction processing on the initial prediction model P1 to obtain the target prediction model P; the prediction module is used to predict the multiple target manufacturing parameters D2 of the inherent strain to be determined using the target prediction model P to obtain the target inherent strain I3 of the molten pool to be determined; the adjustment module is used to complete the adjustment of the target process parameter D4 in the multiple target manufacturing parameters D2 according to the evaluation result of the target inherent strain I3.

[0085] The dimensionless module includes a dimensionless submodule. The dimensionless submodule is used to select the initial manufacturing parameter D1 from the multiple initial manufacturing parameters D1 as the target manufacturing parameter D2; use the target manufacturing parameter D2 as a characteristic parameter to dimensionlessly transform the initial inherent strain I1 to obtain the dimensionless inherent strain I2, which is used to evaluate the residual stress and deformation effect of the formed part; select the target parameter combination D3 from the multiple initial manufacturing parameters D1, wherein the number of the target parameter combinations D3 is the same as the number of types of the dimensionless inherent strain I2, and the target parameter combination D3 corresponds to the type of the dimensionless inherent strain I2.

[0086] The building module includes a building submodule and a dimensionality reduction submodule.

[0087] Among them, the construction submodule is used to determine the number of the target parameter combinations D3 according to the number of the dimensionless inherent strains I2; when the number of the target parameter combinations D3 is multiple, the multiple target parameter combinations D3 are determined as inputs and the multiple dimensionless inherent strains I2 are outputs, and multiple initial prediction models P1 are constructed.

[0088] The dimension reduction submodule is used to determine the number N of dimensionless number groups corresponding to the target parameter combination D3 according to the independent dimensions and independent variables corresponding to the multiple target production parameters D2 in the target parameter combination D3; use the active subspace dimensionality reduction method to obtain multiple target dimensionless number groups G sorted by importance; and construct the target prediction model P based on the multiple target dimensionless number groups G.

[0089] The dimensionality reduction submodule includes: a construction unit, a first determination unit and a second determination unit, the construction unit is used to construct an initial polynomial model P2; determine the multiple initial dimensionless number groups G1 as the input of the initial polynomial model P2, and determine the dimensionless inherent strain I2 as the output of the initial polynomial model P2.

[0090] Among them, the first determination unit is used to determine the multiple target dimensionless number groups G as input, the dimensionless inherent strain I2 as output, and construct the target prediction model P. The initial polynomial model P2 is trained by the polynomial ridge regression method, and the cross-validation method is used to take the determination coefficient as an indicator, until the determination coefficient corresponding to the initial polynomial model P2 is the largest, and then the target order of the polynomial and the multiple target dimensionless number groups G retained in the multiple initial dimensionless number groups G1 are determined; the target prediction model P is constructed based on the multiple target dimensionless number groups G.

[0091] The second determining unit determines the number N1 of the plurality of independent dimensions; determines the number N2 of the plurality of independent variables; and determines the number N of the plurality of dimensionless number groups by the difference between the N2 and the N1.

[0092] Embodiment 3: According to another aspect of the embodiments of the present application, a non-volatile storage medium is also provided, in which a program is stored, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute the above-mentioned method for predicting inherent strain of metal additive manufacturing.

[0093] Embodiment 4: According to another aspect of the embodiments of the present application, there is also provided a computer device, comprising: a memory and a processor, the processor being used to run a program stored in the memory, wherein the above-mentioned data-driven method for predicting inherent strain in metal additive manufacturing is executed when the program is running.

[0094] like Figure 3 As shown, the computer device 10 or the mobile device 10 may include one or more processors 102, a memory 104 for storing data, and a transmission module 106 for communication functions. The processor 102 in the figure is shown as 102a, 102b, ..., 102n, and the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA.

[0095] In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. A person skilled in the art will understand that Figure 3 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 3 More or fewer components as shown, or with Figure 3 Different configurations are shown.

[0096] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuits". The data processing circuits may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuit may be a single independent processing module, or may be incorporated in whole or in part into any of the other components in the computer terminal 10 (or mobile device). As involved in the embodiments of the present application, the data processing circuit acts as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0097] The memory 104 may be used to store software programs and modules of application software, such as a program instruction data storage device corresponding to the method for predicting inherent strain in metal additive manufacturing in the embodiment of the present application.

[0098] The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, realizing the above-mentioned prediction method of inherent strain of metal additive manufacturing. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0099] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer terminal 10. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0100] The display may be, for example, a touch screen liquid crystal display (LCD) that enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0101] In the present invention, the symbols are defined as follows:

[0102] D1: initial manufacturing parameters; D2: target manufacturing parameters; D3: target parameter combination; D4: target process parameters;

[0103] P1: initial prediction model; P2: initial polynomial model; P: target prediction model;

[0104] I1: initial intrinsic strain; I2 dimensionless intrinsic strain; I3: target intrinsic strain;

[0105] N1: number of independent dimensions; N2: number of independent variables; N: number of dimensionless groups;

[0106] G1: initial dimensionless number group; G: target dimensionless number group.

[0107] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for predicting inherent strain in metal additive manufacturing, characterized in that: The following steps are involved: S1. Obtaining initial production parameters D1: The initial production parameters D1 include various process parameters and material property parameters; S2. Obtaining a target parameter combination D3: selecting a target manufacturing parameter D2 from the initial manufacturing parameter D1, and using the target manufacturing parameter D2 to dimensionlessly transform the initial inherent strain I1 to obtain the dimensionless inherent strain I2, and determining a target parameter combination D3 corresponding to the dimensionless inherent strain I2 in the initial manufacturing parameter D1; S3. Obtaining the target prediction model P: Taking the target parameter combination D3 as input and the dimensionless inherent strain I2 as output, constructing the initial prediction model P1, and performing dimensionality reduction processing on the initial prediction model P1 to obtain the target prediction model P. The target prediction model P is shown in formulas (1)-(3): （1）， （2）， （3）， In the formula, , , represent the target dimensionless number groups related to the inherent strains in three directions respectively; yes The polynomial form of yes The polynomial form of yes The polynomial form of ; express The regression prediction model coefficients are express Regression prediction model coefficients, express Regression prediction model coefficients; Among them, the determination coefficient R of the prediction model P 2 The calculation formula of is shown in formula (4). （4）， In the formula, represents the mean of the true output of the model, Represents the model output variable The predicted value of represents the total residual sum of squares, represents the total sum of squares of deviations; S4. Obtaining the target inherent strain I3: The target production parameter D2 of the inherent strain to be determined is predicted using the target prediction model P to obtain the target inherent strain I3; S5. According to the evaluation result of the target inherent strain I3, the target process parameter D4 in the multiple target manufacturing parameters D2 is adjusted.

2. A method for predicting inherent strain in metal additive manufacturing according to claim 1, characterized in that: The step S2 includes: the initial production parameter D1 includes a plurality of parameters, and a parameter is selected from the plurality of initial production parameters D1 as a target production parameter D2; The target manufacturing parameter D2 is used as a characteristic parameter, and the initial inherent strain I1 is dimensionless to obtain the dimensionless inherent strain I2, and the dimensionless inherent strain I2 is used to evaluate the residual stress and deformation of the formed part; The target parameter combination D3 is selected from the multiple initial production parameters D1, wherein the number of the target parameter combinations D3 is the same as the type and quantity of the dimensionless inherent strain I2, and the target parameter combination D3 corresponds to the type of the dimensionless inherent strain I2 one by one.

3. A method for predicting inherent strain in metal additive manufacturing according to claim 2, characterized in that: In the step S3, the number of the target parameter combinations D3 is determined according to the number of the dimensionless inherent strains I2; Multiple one-to-one corresponding target parameter combinations D3 and dimensionless inherent strain I2 are used as input and output respectively to construct multiple initial prediction models P1; Determine the number N of initial dimensionless number groups corresponding to the target parameter combination D3 according to the independent dimensions and independent variables corresponding to the multiple target parameters in the target parameter combination D3; The active subspace dimensionality reduction method is used to reduce the dimension of the initial prediction model P1, and multiple initial dimensionless number groups G1 are obtained in order of importance. The target prediction model P is constructed based on the multiple initial dimensionless number groups G1.

4. A method for predicting inherent strain in metal additive manufacturing according to claim 3, characterized in that: The method for constructing the target prediction model P based on multiple initial dimensionless number groups G1 specifically includes: Construct the initial polynomial model P2; Determine the multiple initial dimensionless number groups G1 as the input of the initial polynomial model P2, determine the dimensionless inherent strain I2 as the output of the initial polynomial model P2, adopt the polynomial ridge regression method to train the initial polynomial model P2, and adopt the cross-validation method to use the determination coefficient as an indicator, until the determination coefficient corresponding to the initial polynomial model P2 is the largest, determine the target order of the polynomial and the multiple target dimensionless number groups G retained in the multiple initial dimensionless number groups G1; The plurality of target dimensionless number groups G are determined as input, the dimensionless inherent strain I2 is determined as output, and the target prediction model P is constructed.

5. The method for predicting inherent strain in metal additive manufacturing according to claim 3, characterized in that: The number N of initial dimensionless number groups corresponding to the target parameter combination D2 is determined by the following method, including: Determine the number N1 of independent dimensions corresponding to the multiple target parameters in the target parameter combination D3; Determine the number N2 of independent variables corresponding to the multiple target parameters in the target parameter combination D3; The difference between N2 and N1 is determined as the number N of the multiple initial dimensionless number groups.

6. A metal additive manufacturing inherent strain prediction system according to a metal additive manufacturing inherent strain prediction method according to any of the above claims, characterized in that: It includes acquisition module, dimensionless module, construction module, prediction module and adjustment module, among which, An acquisition module, used to acquire initial manufacturing parameters D1 of additive manufacturing, wherein the initial manufacturing parameters D1 include process parameters and material property parameters; A dimensionless conversion module, used for selecting a target manufacturing parameter D2 from the multiple initial manufacturing parameters D1, and performing dimensionless conversion of the inherent strain I1 using the target manufacturing parameter D2 to obtain the dimensionless inherent strain I2, and determining a target parameter combination D3 corresponding to the dimensionless inherent strain I2 among the multiple manufacturing parameters D2; A construction module, used to construct an initial prediction model P1 with the target parameter combination D3 as input and the dimensionless inherent strain I2 as output, and perform dimensionality reduction processing on the initial prediction model P1 to obtain a target prediction model P; A prediction module, used to predict a plurality of target manufacturing parameters D2 of the to-be-determined inherent strain using the target prediction model P to obtain a target inherent strain I3 of the to-be-determined molten pool; The adjustment module is used to adjust the target process parameter D4 among the multiple target manufacturing parameters D2 according to the evaluation result of the target inherent strain I3.

7. The metal additive manufacturing inherent strain prediction system according to claim 6, characterized in that: The construction module includes a construction submodule and a dimension reduction submodule, wherein the construction submodule is used to determine the number of the target parameter combinations D3 according to the number of the dimensionless inherent strains I2; when there are multiple target parameter combinations D3, the multiple target parameter combinations D3 are respectively determined as inputs, and the multiple dimensionless inherent strains I2 are outputs, so as to construct multiple initial prediction models P1; The dimension reduction submodule is used to determine the number N of dimensionless number groups corresponding to the target parameter combination D3 according to the independent dimensions and independent variables corresponding to the multiple target production parameters D2 in the target parameter combination D3; and to obtain multiple target dimensionless number groups G sorted by importance using an active subspace dimension reduction method; The target prediction model P is constructed based on the multiple target dimensionless number groups G.

8. The metal additive manufacturing inherent strain prediction system according to claim 7, characterized in that: The dimension reduction submodule includes a construction unit, a first determination unit and a second determination unit. The construction unit is used to construct an initial polynomial model P2, determine a plurality of initial dimensionless number groups G1 as inputs of the initial polynomial model P2, and determine the dimensionless inherent strain I2 as output of the initial polynomial model P2; The first determination unit is used to determine the plurality of target dimensionless number groups G as input, the dimensionless inherent strain I2 as output, and construct the target prediction model P; The second determination unit determines the number N1 of the plurality of independent dimensions; determines the number N2 of the plurality of independent variables; and determines the difference between N2 and N1 as the number N of the plurality of dimensionless number groups.

9. A non-volatile storage medium, characterized in that: The non-volatile storage medium stores a program, wherein when the program is running, the device where the non-volatile storage medium is located is controlled to execute a method for predicting inherent strain in metal additive manufacturing as described in any one of claims 1 to 5.

10. A computer device, characterized in that: include: A memory and a processor, wherein the processor is used to run a program stored in the memory, wherein when the program is run, a method for predicting inherent strain in metal additive manufacturing as described in any one of claims 1 to 5 is executed.

Images