Non-equiaxial residual stress prediction method based on indentation technology
By constructing a mapping relationship between indentation characteristics and material mechanical properties and residual stress, the detection difficulties of existing indentation technology in complex scenarios are solved, and high-precision non-destructive testing is achieved, which is suitable for aerospace and high-end equipment manufacturing fields.
Patent Information
- Application Number
- CN202510948140.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-10
- Publication Date
- 2025-10-10
AI Technical Summary
When detecting non-equiaxed residual stress, existing indentation technology requires known material mechanical properties parameters and a stress-free specimen benchmark, which makes it difficult to adapt to complex stress states and limits its application in complex scenarios.
By constructing a technical system of finite element simulation of multi-indentation process-multi-parameter database-deep learning mapping, a mapping relationship between indentation characteristics and material mechanical properties and residual stress is established, realizing high-precision non-destructive testing without the need for stress-free specimen reference and material mechanical properties parameters.
It achieves efficient and accurate detection of unknown materials and complex stress fields, lowers the threshold for engineering applications, and ensures the safety of component service.
Smart Images
Figure CN120764100A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of material residual stress detection, and in particular relates to a non-equiaxed residual stress prediction method based on indentation technology. Background Art
[0002] Residual stress is the internal stress generated during material processing and manufacturing due to uneven temperature fields and differential deformation. It is divided into residual compressive stress and tensile stress. Residual compressive stress can be used to create a "protective layer" on the component surface through shot peening and rolling processes, inhibiting fatigue cracks. However, residual tensile stress significantly weakens the material's toughness and accelerates damage. When these two factors act in concert to exert external loads, they can easily lead to stress concentration, performance degradation, and even failure in the component. Accurately measuring residual stress is crucial to ensuring component safety in service.
[0003] Residual stress testing technology has been developed to date and is divided into destructive and non-destructive measurement methods based on whether the test damages the material. Although destructive measurement methods (drilling, cutting, etc.) are mature in theory, they will damage the material and weaken the performance; non-destructive measurement methods (XRD, neutron diffraction, etc.) do not damage the material, but have limitations such as low spatial resolution, harsh testing conditions, and high costs, and an efficient and unified testing standard has not yet been formed. Indentation testing technology has developed rapidly due to its advantages of both non-destructive testing and simultaneous acquisition of multiple parameters. It can simultaneously detect mechanical performance parameters such as residual stress, hardness, and elastic modulus, and is easy to operate. However, when existing indentation methods are used to detect non-equiaxed residual stress, they face problems such as insufficient universality of theoretical models, reliance on stress-free specimen benchmarks, and the need for known material mechanical performance parameters, which limits their application in complex scenarios.
[0004] Therefore, the present invention provides a non-equiaxed residual stress prediction method based on indentation technology, which breaks through the limitations of existing indentation technology and does not require known material mechanical properties and stress-free sample references. Through finite element simulation of elastic-plastic materials under non-equiaxed residual stress, the indentation process of flat indenter and ball indenter at the same position and the indentation process of ball at a new position are constructed to construct a multi-parameter, multi-stress level indentation simulation database; based on the database, a mapping relationship between indentation characteristics and material mechanical properties and residual stress is constructed, and a non-equiaxed residual stress prediction model is established to achieve high-precision prediction of non-equiaxed residual stress, providing efficient non-destructive testing technology for aerospace, high-end equipment manufacturing and other fields, and ensuring the safe and reliable service of components. Summary of the Invention
[0005] The purpose of the present invention is to provide a non-equiaxed residual stress prediction method based on indentation technology, breaking through the limitations of existing indentation detection technology. In view of the problems that the traditional indentation method needs to know the mechanical properties of the material in advance, relies on the stress-free specimen benchmark, and the theoretical model is difficult to adapt to complex stress states when detecting non-equiaxed residual stress, by constructing an innovative technical system of "multi-indentation process finite element simulation-multi-parameter database construction-deep learning mapping", high-precision non-destructive detection of non-equiaxed residual stress without the need for stress-free specimen reference and preset material mechanical properties parameters is achieved, which greatly improves the efficiency and accuracy of residual stress detection, provides a universal residual stress detection technology solution for aerospace, high-end equipment manufacturing and other fields, and effectively promotes the intelligent and efficient development of complex stress detection technology.
[0006] The present invention provides a non-equiaxed residual stress prediction method based on indentation technology, which specifically includes the following steps:
[0007] Step 1: Establish a finite element model of material indentation under non-equiaxed residual stress, generate multiple combinations of material mechanical property parameters based on the Latin hypercube sampling method, and simulate the indentation process of elastic-plastic materials at the same position of flat indenter and spherical indenter;
[0008] Step 2: Based on the finite element simulation of two indentations at the same location with various parameter combinations, a database of indentation characteristics and material mechanical parameters is established;
[0009] Step 3: Based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and material mechanical properties parameters, and the material mechanical properties parameters are calculated;
[0010] Step 4: The calculated material mechanical property parameters and the multiple residual stress parameters generated based on the Latin hypercube sampling method are combined and input into the finite element model to perform a single ball indentation test at a new position of the material under the action of non-equiaxed residual stress;
[0011] Step 5: Based on the indentation test of the ball indenter at the new position of the material, a database of indentation characteristics and non-equiaxed residual stress is established;
[0012] Step 6: Based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and residual stress, and a non-equiaxed residual stress prediction model is established.
[0013] Furthermore, in step 1, a finite element model of material indentation under non-equiaxed residual stress is established, including: using ABAQUS software; the design size is 0.4×0.4×0.4mm 3a three-dimensional 1 / 4 symmetry finite element model sample; the model adopts a uniform grid structure, the grid near the center contact area of the indentation is refined, the grid density gradually decreases away from the center contact area, the number of units is determined through grid convergence verification, the model sample has a total of 80275 eight-node linear hexahedral units (C3D8R); the boundary conditions are to restrict the Z-direction translation and X, Y-direction rotation of the nodes on the XZ face of the bottom of the specimen, to restrict the Y-direction translation of the nodes on the XZ symmetry plane, and to restrict the X-direction translation of the nodes on the YZ symmetry plane; the stresses σ R-X and σ R-Z are simulated by setting an initial stress field in the model (“+” represents tensile stress, and “-” represents compressive stress); the indenter material is diamond, which is sufficiently hard relative to the material, and is set as a rigid body, and the friction coefficient between the indenter and the specimen is set to 0.25; the model uses two kinds of indenters and indentation control conditions (suitable for indentation models of the two kinds of indenters), and all other settings of the finite element model are the same except for the difference in the indenters, the flat indenter diameter D f = 0.1 mm, a flat indenter round corner transition is set to prevent stress concentration after indentation, the ratio of the flat indenter diameter to the ball indenter diameter D f / D s = 1 is fixed; the starting position of the ball indenter indentation is set to be the same as that of the flat indenter, and the indentation depths of the flat indenter and the ball indenter are set to 5 μm.
[0014] Further, in step 1, a plurality of material mechanical property parameter combinations are generated based on the Latin hypercube sampling method, the material mechanical property parameter combinations include an elastic modulus E, a yield strength σ y and a hardening index n, and the parameter ranges include an elastic modulus E of 50-150 GPa, a yield stress σ y of 200-1000 MPa, and a hardening index n of 0.05-0.20.
[0015] Further, in step 2, a database of indentation characteristics and material mechanical parameters is established based on finite element simulation of two times of indentation at the same position for each parameter combination, including: analyzing the load-displacement curve obtained after the second step of the ball indentation at the same position, selecting the fitting parameters in the loading stage and the unloading stage of the load-displacement curve, the indentation work and the indentation displacement as analysis parameters, and establishing an indentation parameter matrix [W e / W p , W e / W t , W p / W t , h f / h max , a, m, C] are all ratio parameters, and the database is established together with the material mechanical parameters.
[0016] Further, in step 3, based on the indentation simulation database, a deep neural network is used to construct the mapping relationship between the indentation features and the material mechanical performance parameters, and the material mechanical performance parameters are calculated, including: taking the indentation features [W e / W p , W e / W t , W p / W t , h f / h max , a, m, C] extracted from the load-displacement curve as the input layer of the deep neural network, and the material performance parameters [E, σ y , n] as the output layer for model training, according to the deep learning principle, the target mapping function is as follows:
[0017] (1)
[0018] The deep neural network is used to establish the nonlinear mapping relationship between the input and the output, the model is trained by dividing the training set, the validation set and the test set, and the network parameters (including but not limited to hidden layer weight, bias) are adjusted by multiple iterations to optimize the mapping precision.
[0019] The trained deep neural network model can output the predicted value of the corresponding material mechanical performance parameter according to the input indentation feature parameter.
[0020] Further, in step 4, the calculated material mechanical performance parameters and the multiple residual stress parameters generated based on the Latin hypercube sampling method are combined to input the finite element model, and the single ball indentation test of the material new position under the action of the anisotropic residual stress is performed, including: combining the material mechanical performance parameters (E, σ y , n) with the anisotropic residual stress parameters generated by the Latin hypercube sampling, and inputting the finite element model to perform the single ball indentation test of the material new position under the action of the anisotropic residual stress; the anisotropic residual stress parameters at least include X-direction and Z-direction residual stress, taking the yield stress σ y as the normalization reference, setting the anisotropic residual stress parameters σ R-X / σ y , σ R-Z / σ y , the value range is -0.9~0.9.
[0021] Further, in step 5, based on the ball indentation test of the material new position, a database of indentation features and anisotropic residual stress is established, including: analyzing the load-displacement curve obtained after the ball indentation of the material new position, selecting the fitting parameters in the loading stage and the unloading stage of the load-displacement curve, the indentation size, the indentation work and the indentation displacement as the analysis parameters, and establishing the indentation parameter matrix [W e / Wp , W e / W t , W p / W t , h f / h max ,a,m,C,r x , r z ] are all ratio parameters, and the database is established together with the non-equiaxed residual stress.
[0022] Furthermore, in step 6, based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and residual stress, and a non-equiaxed residual stress prediction model is established, including: designing the X-direction residual stress and the Z-direction residual stress [σ R-X / σ y , σ R-Z / σ y ] parameter combination, the indentation features extracted from the load-displacement curve [W e / W p , W e / W t , W p / W t , h f / h max ,a,m,C,r x , r z ] as the input layer of the deep neural network, the material residual stress [σ R-X / σ y , σ R-Z / σ y ] is used as the output layer for model training. According to the principle of deep learning, the target mapping function is as follows:
[0023] (2)
[0024] A deep neural network is used to establish a nonlinear mapping relationship between input and output. Model training is performed by dividing the model into training sets, validation sets, and test sets. Network parameters (including but not limited to hidden layer weights and biases) are adjusted through multiple rounds of iterations to optimize mapping accuracy.
[0025] The trained deep neural network model can output the corresponding material residual stress parameter prediction value based on the input indentation characteristic parameters.
[0026] Beneficial effects of the present invention:
[0027] This method, based on finite element simulation technology for multiple indentation processes, constructs a dual-dimensional database encompassing material mechanical properties and non-equiaxed residual stresses, completely eliminating reliance on stress-free specimen benchmarks and known material performance parameters. Without the need for complex specimen preparation or performance prediction testing, it can directly test in-service components and components with unknown material properties, significantly lowering the barrier to engineering application and addressing the challenges of traditional methods in testing unknown materials and complex stress fields.
[0028] In addition, through the collaborative optimization of deep neural networks and Latin hypercube sampling method, the nonlinear mapping relationship between indentation characteristics and residual stress can be accurately captured. The obtained material mechanical properties parameters and non-equiaxed residual stress prediction results can provide key data support for component process optimization and service life assessment.
[0029] Therefore, the non-equiaxed residual stress prediction method based on indentation technology proposed in the present invention can be promoted and applied as a more universal detection technology, providing efficient and non-destructive residual stress detection solutions for high-end manufacturing fields such as aerospace and new energy equipment, and effectively ensuring the service safety and reliability of key components. BRIEF DESCRIPTION OF THE DRAWINGS
[0030] The drawings described herein are used to provide further understanding of the present invention and constitute a part of this application. The illustrative examples of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.
[0031] Figure 1 It is a flow chart of the non-equiaxed residual stress prediction method based on indentation technology described in the present invention.
[0032] Figure 2 It is a flow chart of the indentation test with a flat indenter and a ball indenter in an embodiment of the present invention.
[0033] Figure 3 It is a cross-sectional view of the indentation morphology after indentation unloading in an embodiment of the present invention. DETAILED DESCRIPTION
[0034] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0035] Example 1
[0036] Combine Figure 1 The present invention provides a non-equiaxed residual stress prediction method based on indentation technology, the steps are as follows:
[0037] Step 1: Establish a finite element model of material indentation under non-equiaxed residual stress. Generate multiple combinations of material mechanical property parameters based on the Latin hypercube sampling method to simulate the indentation process of elastic-plastic materials at the same position of a flat indenter and a spherical indenter. The details are as follows:
[0038] Establish a finite element model of material indentation under non-equiaxed residual stress: use ABAQUS software; the model adopts mm, MPa, t / mm3 units, and the design size is 0.4×0.4×0.4mm3 of a three-dimensional 1 / 4 symmetrical finite element model specimen; the model adopts a uniform grid structure, the grid near the indentation center contact area is refined, and the grid density gradually decreases away from the center contact area. The number of units is determined by grid convergence verification. The model specimen has a total of 80275 eight-node linear hexahedral units (C3D8R); the boundary conditions are to limit the Z-direction translation and X and Y-direction rotation of the XZ plane nodes at the bottom of the specimen, limit the Y-direction translation of the XZ symmetry plane nodes, and limit the X-direction translation of the YZ symmetry plane nodes; by setting the initial stress field in the model, σ R-X and σ R-Z Stress ("+" represents tensile stress, "-" represents compressive stress) simulates the equivalent non-equiaxed residual stress in the specimen; the indenter is made of diamond, which is hard enough relative to the material and is set as a rigid body. The friction coefficient between the indenter and the specimen is set to 0.25; the model uses two different indenters and indentation control conditions (applicable to the indentation model of the two indenters). Except for the different indenters, all other settings of the finite element model are the same. The flat indenter diameter D f =0.1mm, in order to prevent stress concentration after indentation, set the flat indenter radius transition, fix the ratio of the flat indenter diameter to the ball indenter diameter D f / D s =1; set the starting position of the ball indenter to be the same as that of the flat indenter, and set the indentation depth of the flat indenter and the ball indenter to 5μm.
[0039] Since the measured friction coefficient between the diamond indenter and the metal material is usually between 0.2 and 0.3, a representative middle value in the commonly used friction coefficient range is selected without introducing significant errors. In this example, the friction coefficient between the indenter and the sample is selected as 0.25. For linear elastic-power hardening materials, small fluctuations in the Poisson's ratio have little effect on the indentation response. The Poisson's ratio of most metal materials is generally concentrated around 0.3. In this example, the Poisson's ratio of the material is uniformly set to 0.3. Set the material mechanical property parameter range E, σ y , n is 50~150GPa, 200~1000MPa, 0.05~0.20 respectively, and the mechanical properties of the material are uniformly taken within this range, that is, the elastic modulus E is 50GPa, 100GPa, 150GPa, and the yield stress σ yTake 200 MPa, 400 MPa, 600 MPa, 800 MPa, 1000 MPa, strain hardening index n takes 0.05, 0.10, 0.15, 0.20.
[0040] In order to express the material performance parameters in the range of material mechanics performance more widely, Latin hypercube sampling method is used to generate various combinations of input material parameters of finite element model. In the range of elastic modulus E of 50~150GPa, yield stress σ y Range of 200~1000MPa, hardening index n range of 0.05~0.20, 200 parameter combinations are generated.
[0041] For each material parameter combination, the same position flat head, ball indenter indentation simulation as Figure 2 Shown in the figure.
[0042] Step 2, based on the finite element simulation of the same position twice indentation of each parameter combination, the database of indentation characteristics and material mechanics parameters is established, as follows:
[0043] According to the load-displacement curve obtained after the second step of the same position ball indentation, the fitting parameters, indentation work and indentation displacement in the loading stage and unloading stage of the load-displacement curve are selected as the analysis parameters, and the indentation parameter matrix [W e / W p , W e / W t , W p / W t , h f / h max , a, m, C] are all ratio parameters, which can perfectly unify the parameter form. h f Is the residual indentation displacement, h max Is the maximum indentation displacement, the change of material indentation work mainly depends on the material mechanics performance parameters and residual stress, wherein the total indentation work is equal to the sum of elastic work and plastic work, and the function expression of indentation elastic work and indentation plastic work is:
[0044] (3)
[0045] (4)
[0046] The obtained indentation characteristic parameter matrix and material mechanics parameters are used to establish the database.
[0047] Step 3, based on the indentation simulation database, the mapping relationship between indentation characteristics and material mechanics performance parameters is constructed by using deep neural network, and the material mechanics performance parameters are calculated, as follows:
[0048] By writing a training program based on MATLAB, introducing a deep neural network, and establishing a mapping relationship between indentation characteristics and material mechanical properties parameters, the indentation characteristics extracted from the load-displacement curve [W e / W p , W e / W t , W p / W t , h f / h max , a, m, C] as the input layer, material performance parameters [E, σ y , n] is used as the output layer for model training. According to the principle of deep learning, the target mapping function is as follows:
[0049] (1)
[0050] Before the training phase, the model parameters are initialized and the dataset is divided into training set, validation set, and test set in a ratio of 6:2:2. The dataset, initial weights, and bias of the hidden layer are different in each training process, and different solutions will be obtained for the same problem. Therefore, the model dataset needs to be trained multiple times for optimization.
[0051] Step 4: Input the calculated material mechanical properties parameters and the multiple residual stress parameters generated based on the Latin hypercube sampling method into the finite element model to perform a single ball indentation test at a new position of the material under the action of non-equiaxed residual stress, as follows:
[0052] The calculated material mechanical properties parameters (E, σ y , n) input into the finite element model, simulate the indentation process under different residual stresses of the material, and set the non-equiaxed residual stress parameter σ R-X / σ y , σ R-Z / σ y The value range is -0.9~0.9. The Latin hypercube sampling method is used to randomly sample and obtain 200 parameter combinations of non-equiaxed residual stress parameters. Then the parameter combination is input into the finite element model to carry out the third step of the single ball indentation finite element simulation of the new material position, such as Figure 2 shown.
[0053] Step 5: Based on the ball indentation test at the new position of the material, a database of indentation characteristics and non-equiaxed residual stress is established, as follows:
[0054] The load-displacement curve obtained after the ball indentation at the new position of the material is analyzed, and the fitting parameters of the load-displacement curve in the loading and unloading stages, indentation size, indentation work and indentation displacement are selected as analysis parameters to establish the indentation parameter matrix [W e / W p , We / W t , W p / W t , h f / h max ,a,m,C,r x , r z ] are all ratio parameters. Figure 3 Shown are E, σ y When n is 100 GPa, 600 MPa and 0.15 respectively, the indentation profiles in the X and Z directions after the non-equiaxed residual stress indentation of the material are completed. The indentation bulge and the asymmetry of the indentation morphology are related to the principal stress in the direction. The long axis of the elliptical indentation indicates the direction of the maximum principal stress, and the short axis appears in the direction of the minimum principal stress. x and r z The two characteristic parameters represent the X-direction and Z-direction indentation marks of the indentation ellipse, and the direction of the principal stress can be directly determined by the asymmetric morphology of the residual indentation.
[0055] The database is established by combining the obtained indentation characteristic parameter matrix with the non-equiaxed residual stress.
[0056] Step 6: Based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and residual stress, and a non-equiaxed residual stress prediction model is established, as follows:
[0057] The training program was written in MATLAB to design the X-axis residual stress and Z-axis residual stress [σ R-X / σ y ,σ R-Z / σ y ] parameter combination, the extracted indentation features [W e / W p , W e / W t , W p / W t , h f / h max ,a,m,C,r x , r z ] as the input layer of the deep neural network, the material residual stress [σ R-X / σ y ,σ R-Z / σ y ] is used as the output layer for model training, and the target mapping function is as follows:
[0058] (2)
[0059] Before the training phase, the model parameters were initialized, the training set, validation set, and test set were divided into 6:2:2 ratios, and the model dataset was trained multiple times to find the optimal value.
[0060] The trained deep neural network model can output the corresponding material residual stress parameter prediction value based on the input indentation characteristic parameters.
[0061] The above content is merely an exemplary description of the present invention and does not constitute a limitation of the present invention. For those skilled in the art, various other forms of changes or modifications may be made based on this description. In view of the diversity of implementation methods, it is impossible to enumerate all situations one by one. However, any modifications, equivalent replacements, improvements, etc. derived from the present invention should be included in the scope of protection of the present invention.
Claims
1. A non-equiaxed residual stress prediction method based on indentation technology, characterized in that: The steps include: Step 1: Establish a finite element model of material indentation under non-equiaxed residual stress, generate multiple combinations of material mechanical property parameters based on the Latin hypercube sampling method, and simulate the indentation process of elastic-plastic materials at the same position of flat indenter and spherical indenter; Step 2: Based on the finite element simulation of two indentations at the same location with various parameter combinations, a database of indentation characteristics and material mechanical parameters is established; Step 3: Based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and material mechanical properties parameters, and the material mechanical properties parameters are calculated; Step 4: The calculated material mechanical property parameters and the multiple residual stress parameters generated based on the Latin hypercube sampling method are combined and input into the finite element model to perform a single ball indentation test at a new position of the material under the action of non-equiaxed residual stress; Step 5: Based on the indentation test of the ball indenter at the new position of the material, a database of indentation characteristics and non-equiaxed residual stress is established; Step 6: Based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and residual stress, and a non-equiaxed residual stress prediction model is established.
2. The non-equiaxed residual stress prediction method based on indentation technology according to claim 1, characterized in that: In step 1, a finite element model of material indentation under non-equiaxed residual stress is established, including: A symmetrical finite element model specimen is used to reduce the computational complexity through reasonable mesh division or model simplification. The boundary conditions are to limit the rigid body displacement of the specimen bottom and the symmetry surface to ensure the rationality of the model mechanical boundary. By setting the initial stress field in the model, σ R-X and σ R-Z Stress is simulated to simulate the equivalent non-equiaxed residual stress in the specimen; two indenters, a flat indenter and a spherical indenter, are used to keep the same indentation starting position. The indenter is set as a rigid body, and stress concentration is avoided by controlling the size parameters (such as diameter ratio) of the flat indenter and the spherical indenter.
3. The non-equiaxed residual stress prediction method based on indentation technology according to claim 1, characterized in that: In step 1, multiple material mechanical property parameter combinations are generated based on the Latin hypercube sampling method, and the material mechanical property parameter combinations include elastic modulus E, yield strength σ y and hardening index n.
4. The non-equiaxed residual stress prediction method based on indentation technology according to claim 1, characterized in that: In step 2, based on the finite element simulation of two indentations at the same location with various parameter combinations, a database of indentation characteristics and material mechanical parameters is established, including: The indentation characteristics after the second step of ball indentation at the same position, including energy ratio and geometric feature ratio, are used to establish a database together with the material mechanical parameters.
5. The non-equiaxed residual stress prediction method based on indentation technology according to claim 1, characterized in that: In step 3, based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and material mechanical properties parameters, and the material mechanical properties parameters are calculated, including: Based on the indentation simulation database, the indentation characteristics are used as input, and the mechanical properties parameters (E, σ y , n) is the output, and the deep neural network is trained to establish a nonlinear mapping relationship between input and output. Through reasonable model training and parameter optimization strategies, the mechanical properties parameters of the material are calculated.
6. The non-equiaxed residual stress prediction method based on indentation technology according to claim 1, characterized in that: In step 4, the calculated material mechanical properties and the various residual stress parameters generated based on the Latin hypercube sampling method are combined and input into the finite element model to perform a single-ball indentation test at a new position of the material under the action of non-equiaxed residual stress, including: The material mechanical properties parameters (E, σ y , n) is combined with the non-equiaxed residual stress parameters generated by the Latin hypercube sampling method and input into the finite element model to perform a single ball indentation test of the new position of the material under the action of non-equiaxed residual stress; the non-equiaxed residual stress parameters at least include the residual stress in the X direction and the Z direction, with the yield stress σ y is the normalized benchmark.
7. The non-equiaxed residual stress prediction method based on indentation technology according to claim 1, characterized in that: In step 5, the ball indenter indentation test is performed based on the new material position to establish a database of indentation characteristics and non-equiaxed residual stresses, including: The indentation characteristics after ball indentation at the new position of the material, including energy ratio and geometric characteristic ratio, are used to establish a database together with the non-equiaxed residual stress.
8. The non-equiaxed residual stress prediction method based on indentation technology according to claim 1, characterized in that: In step 6, based on the indentation simulation database, a deep neural network is used to construct a mapping relationship between indentation characteristics and residual stress, and a non-equiaxed residual stress prediction model is established, including: Based on the indentation simulation database, with indentation characteristics as input and material residual stress as output, a deep neural network is trained to establish a nonlinear mapping relationship between input and output. Through a reasonable model training strategy, a non-equiaxed residual stress prediction model is established. The model with the smallest error in multiple training runs is used as the final prediction model.