Residual stress region reconstruction method and system

Abstract

The invention belongs to the field of residual stress reconstruction control, and particularly relates to a residual stress region reconstruction method and system. According to the method, residual stress distribution after laser forging treatment is reconstructed by combining an iteration method based on limited measurement point least square fitting, firstly, the iteration method is started by introducing a specified stress component into a data measurement area of a finite element model, and then static balance analysis is carried out; if the stress distribution of the mapping area and the non-mapping area is not converged to the target stress, the stress is redistributed. The input stress is repeatedly introduced into the mapping region, so that the input stress is converged to a specified stress component, and complete residual stress distribution in the whole region can be obtained. When all stress sample points are in a measurement area, the whole residual stress field can be well reconstructed by using an iteration method.

Description

Technical Field

[0001] The present invention belongs to the field of residual stress reconstruction control, and particularly relates to a method and system for reconstructing a residual stress region. Background Art

[0002] Offshore wind power is a clean and renewable energy source. Vigorously developing the offshore wind power industry is an important way to achieve the "dual carbon" goal. Therefore, a large number of wind power-related equipment are entering the marine environment for service. However, the complex and harsh marine environment is extremely likely to cause damage such as wear, corrosion, and cracking to offshore wind power equipment, seriously affecting the service life of the equipment and even causing safety accidents. Accurately and quickly evaluating the residual stress of offshore wind power equipment can effectively control the stress state of the equipment during service or repair, thereby ensuring the operating efficiency of the equipment. Intelligent and efficient stress testing equipment has become an inevitable trend for on-site inspection of offshore wind power equipment.

[0003] For the residual stress region reconstruction system software, the obtained residual stress values are discrete points during the actual stress measurement process. It is necessary to reconstruct the full-field residual stress based on limited measurement data points. When using the inherent strain method to reconstruct the residual stress, there is a problem of inverse solution. By using a numerical iteration method based on the inherent strain theory and multi-dimensional interpolation fitting, the reconstruction of the full-field residual stress distribution based on a small number of stress data points in the stress mapping area is realized, providing an effective means for quickly and accurately reconstructing the residual stress on-site. Summary of the Invention

[0004] The purpose of the present invention is to propose a method and system for reconstructing a residual stress region to solve one or more technical problems existing in the prior art, and at least provide a beneficial choice or create conditions.

[0005] A method for reconstructing a residual stress region, characterized in that the method includes the following steps:

[0006] Read the on-site residual stress test data;

[0007] Manually input the initial unknown parameters, calculate the average relative error threshold through the genetic algorithm, and record the number of runs of the genetic algorithm;

[0008] Through the on-site residual stress test data, realize the reconstruction of the regional residual stress through iterative calculation, and finally output the finite element simulation file of each iteration, the average residual stress value of the corresponding point, and the number of iterations at which the minimum error occurs, and output it in the form of a chart.

[0009] Further, the main steps for testing the data of the residual stress of data reading are as follows: obtaining the residual stress distribution of the repaired component at a certain key cross-section position or multiple positions by X-ray diffraction method; establishing a three-dimensional model with the same geometric dimensions as the component in finite element software, generating finite element meshes, extracting the grid node coordinate information into scientific computing software for numerical fitting processing, and initializing the parameters of the region where the eigenstrain is located.

[0010] Further, the artificially input initial unknown parameter is the initialized eigenstrain. Using an optimization algorithm based on the artificial intelligence genetic algorithm and the fuzzy finite element theory to call the finite element software to perform batch calculations for reconstructing the residual stress according to the eigenstrain, taking the mean square error MSE between the measured residual stress value at the cross-section point and the finite element calculated value as the fitness function, and obtaining the population with the minimum fitness through iterative calculations, that is, obtaining the optimal parameters of the region where the eigenstrain is located; by applying the calculated eigenstrain distribution in the finite element model and performing static analysis, the full-field residual stress distribution with the minimum error from the actually measured residual stress on the cross-section can be obtained.

[0011] Preferably, the eigenstrain theory is often used to reconstruct the residual stress field after laser shock. Eigenstrain is any permanent strain generated during the inelastic deformation of materials, such as plastic deformation, crystallographic transformation, thermal expansion mismatch between different components, etc. In order to offset the deformation incoordination of the eigenstrain, there must be an additional residual elastic strain term in the material deformation compatibility equation to maintain balance, and the stress caused by these residual elastic strains is the residual stress field of the material. Therefore, if the stress distribution in the region containing the eigenstrain is known, the residual term can be expressed without introducing the eigenstrain distribution in the whole solid, and the complete residual stress field can be obtained according to the deformation compatibility equation and the stress balance equation, which is the basic principle of the stress iteration method proposed in this software.

[0012] Preferably, except for the initial measurement process, after the corresponding program is compiled, this method can realize automated batch calculations, without repeated modeling and adjusting optimization parameters, greatly reducing the calculation time, and using the artificial intelligence genetic algorithm can make the optimal parameters approach the global optimal solution, improving the solution accuracy, and can be used to obtain the full-field residual stress distribution of the repaired component when the known finite discrete points are available.

[0013] Further, measure the region where the stress values of the sample points in the region containing all eigenstrains are located from the experiment, define this region as the mapping region, and define the region without eigenstrain as the non-mapping region. Obtain the stress components at each position in the mapping region in the reconstructed finite element model through multi-dimensional least squares fitting, and assign the stress components to the Gauss integration points of the elements in the finite element model to start the iterative process;

[0014] Define the stress σ measured in the mapped area input , and solve a one-step static equilibrium in the first iteration. Due to the introduction of the stress field, the stress components will redistribute in the mapped area and the non-mapped area;

[0015] Define σ ouput as the redistributed stress in the mapped area after the static equilibrium step, and judge whether there is a deviation between σ input and σ ouput . If there is a deviation between σ input and σ ouput , it is determined as the redistributed stress in the mapped area after the static equilibrium step. If there is no deviation between σ input and σ ouput , it is determined that no stress redistribution occurs;

[0016] Take the mean square error between the reconstructed residual stress value σ ouput in the mapped area and the measured residual stress value σ input of the component as the fitness function, and iteratively calculate the population with the minimum fitness. The mean square error formula:

[0017]

[0018] where n is the number of data points, Y i is the i-th actual observed value, is the i-th predicted value;

[0019] Furthermore, to ensure that σ ouput finally converges to σ input after several iterations, the proportional-integral adjustment method is used to modify the output stress to its target measured value during subsequent iterations. The adjustment equation can be expressed as:

[0020]

[0021] where the superscript i represents the i-th iteration, β is the integral factor, and β takes the value of 1, is the initial input stress tensor, that is, the measured convergence value; after the first iteration, the input stress and output stress can be obtained for the next iteration. Repeat to introduce σ input into the mapped area, while the rest of the entire area remains unchanged until and σ input the tolerance between them is less than the predetermined value, so that the full-field stress distribution can be obtained., Execute the implementation procedure of finite element analysis using finite element software, post-process the results of the previous iteration using a script written in parametric design language, and write the input file for the next iteration. During the entire iteration process, the mesh remains unchanged, and the stress components are assigned to the Gaussian integration points of the elements.

[0022] Furthermore, a residual stress area reconstruction system, the overall framework of the system includes: a user terminal, an interactive interface, and a main program. The user terminal is the input port for residual stress test data at the laser forging repair site. The input data includes: network node data, simulation file models, and measured stress data. The interactive interface is the data transmission interface between the user terminal and the main program. The main program includes an interactive system main program and a stress reconstruction main program. The main program can execute a computer program to implement a residual stress area reconstruction method described in any of the above methods.

[0023] The stress reconstruction main program includes the operation of multiple modules, including: an evaluation fitness module, a main program operation module, a running simulation module, a creating or loading neural network model module, a data processing class module, and a neural network and standardization module;

[0024] The functions of the modules are as follows:

[0025] Evaluation fitness module: Calculate the fitness of individuals in the genetic algorithm. Specifically, calculate the relative error here and predict the error through a neural network;

[0026] Main program operation module: The main program executes the loop of the genetic algorithm, including initializing the population, evaluating the fitness of individuals, performing selection, crossover, and mutation operations, and checking whether the termination conditions are met (such as reaching the maximum number of generations or the error being lower than the threshold);

[0027] Running simulation module: Call the simulation software and output the results to a specific file for further error calculation;

[0028] Creating or loading neural network model module: Create a new neural network model or load an existing model according to the settings to provide an optimization direction for the genetic algorithm;

[0029] Data processing class module: Responsible for calculating the stress distribution, which is the mathematical model used by the main program to calculate stress, updating the stress data, and providing an evaluation basis for the genetic algorithm;

[0030] Neural network and standardization module: Manage the creation and loading of neural networks, assist in error prediction and model training, provide standardization support for the training data of neural networks, and ensure the stability of the data;

[0031] The beneficial effects of the present invention are as follows: The present invention reconstructs the residual stress distribution after laser forging treatment through an iterative method based on least squares fitting of finite measurement points. First, the iterative method is initiated by introducing specified stress components into the data measurement area (the mapped area) of the finite element (FE) model, and then a static equilibrium analysis is performed. If the stress distributions in the mapped area and the unmapped area do not converge to the target stress, stress redistribution will occur. By repeatedly introducing input stress into the mapped area to make it converge to the specified stress components, a complete residual stress distribution throughout the area can be obtained. When all stress sample points are within the measurement area, the iterative method used can better reconstruct the entire residual stress field. Moreover, without inversely solving the distribution of inherent strain in the material, the experimental measurement stress points are directly used for reconstruction, which has more advantages in complex geometric structures. In addition, the iterative method adopted by the software is a posteriori method, and its reconstruction performance depends on the accuracy of the experimental measurement points. The accuracy of stress during the reconstruction process can also be ensured by manually adjusting genetic algorithm parameters such as the number of iterations and the threshold of the average relative error. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] To further explain the present invention in detail, the following drawings are provided in this solution and are explained with specific embodiments. The drawings and embodiments do not limit the present invention. The drawings are as follows:

[0033] Figure 1 It is a flowchart of a method for reconstructing a residual stress area;

[0034] Figure 2 It is a technical roadmap for residual stress reconstruction of a method for reconstructing a residual stress area;

[0035] Figure 3 It is a flowchart of stress iteration of a method for reconstructing a residual stress area;

[0036] Figure 4 It is an overall structure diagram of a system for reconstructing a residual stress area;

[0037] Figure 5 It is a software interface of a system for reconstructing a residual stress area;

[0038] Figure 6 It is a graph of the number of iterations - error generated by a system for reconstructing a residual stress area. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0039] The following will clearly and completely describe the concept, specific structure, and technical effects generated by the present invention in combination with the embodiments and the drawings to fully understand the purpose, solution, and effects of the present invention. It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other.

[0040] Laser forging additive repair technology refers to the process of simultaneously and synergistically repairing metal components with an electric arc and a laser beam. It can cause plastic deformation of the cladding layer, eliminate internal defects such as pores and cracks in the deposited layer, and thermal stress, enabling the repaired component to obtain excellent internal quality and comprehensive mechanical properties without subsequent heat treatment, and effectively controlling macroscopic deformation and cracking problems.

[0041] During the high-energy beam laser processing, the material may be subjected to non-uniform force / heat or thermo-mechanical coupling fields in the time domain and spatial domain, resulting in non-uniform deformation and residual stress harmful to fatigue life. Therefore, the research on the deformation mechanism and the formation mechanism of residual stress during the processing can provide theoretical reference and technical support for the further development of laser processing. At the same time, to further study the low-stress long-life on-site repair technology, better understand, predict and control the internal stress generated in the manufacturing, processing or service process of the material structure, improve product quality, performance and service life, it is necessary to reconstruct the full-field residual stress based on limited measurement data points during the actual stress measurement process. Thus, it is of great significance to develop an intelligent residual stress region reconstruction system.

[0042] As Figure 1 shown, a method for reconstructing a residual stress region is characterized in that the method comprises the following steps:

[0043] Read the on-site residual stress test data;

[0044] Manually input the initial unknown parameters, calculate the average relative error threshold through the genetic algorithm, and record the number of times the genetic algorithm runs;

[0045] Realize the regional residual stress reconstruction through iterative calculation with the on-site residual stress test data, and finally output the finite element simulation file for each iteration, the average residual stress value of the corresponding point and the number of iterations at which the minimum error occurs, and output it in the form of a chart.

[0046] Further, the main steps of the data reading the residual stress test data are: obtaining the residual stress distribution at a key cross-section position or multiple positions of the repaired component by X-ray diffraction method; establishing a three-dimensional model with the same geometric dimensions as the component in the finite element software, generating finite element meshes, extracting the grid node coordinate information into the scientific computing software for numerical fitting processing, and initializing the parameters of the region where the inherent strain is located.

[0047] Further, as Figure 2As shown, the initially unknown parameters input manually are the initialized inherent strains. An optimization algorithm based on the artificial intelligence genetic algorithm and the fuzzy finite element theory are used to call the finite element software to perform batch calculations for reconstructing the residual stress according to the inherent strains. The mean square error MSE between the measured residual stress value at the cross-section points and the finite element calculated value is used as the fitness function. Through iterative calculations, the population with the minimum fitness is obtained, that is, the optimal parameters in the area where the inherent strains are located are obtained. By applying the calculated inherent strain distribution in the finite element model and performing static analysis, the full-field residual stress distribution with the minimum error from the actually measured residual stress on the cross-section is obtained.

[0048] Preferably, the inherent strain theory is often used to reconstruct the residual stress field after laser shock. Inherent strain is any permanent strain generated during the inelastic deformation process of materials, such as plastic deformation, crystallographic transformation, thermal expansion mismatch between different components, etc. In order to offset the deformation incoordination of the inherent strain, there must be an additional residual elastic strain term in the material deformation coordination equation to maintain balance. The stress caused by these residual elastic strains is the residual stress field of the material. Therefore, if the stress distribution in the area containing the inherent strain is known, the residual term can be expressed without introducing the inherent strain distribution in the whole solid, and the complete residual stress field can be obtained according to the deformation coordination equation and the stress balance equation. This is the basic principle of the stress iteration method proposed in this software.

[0049] Preferably, except for the initial measurement process, after the corresponding program is compiled, this method can realize automated batch calculations, without repeated modeling and adjustment of optimization parameters, greatly reducing the calculation time. And using the artificial intelligence genetic algorithm can make the optimal parameters approach the global optimal solution, improving the solution accuracy, and can be used to obtain the full-field residual stress distribution of the repaired component when the known finite discrete points are available.

[0050] Furthermore, as Figure 3 shown, measure the area where the stress values of the sample points in the area containing all inherent strains are located in the experiment. Define this area as the mapping area, and the area not containing the inherent strain is defined as the non-mapping area. Obtain the stress components at each position in the mapping area in the reconstructed finite element model through multi-dimensional least squares fitting. The stress components are assigned to the Gauss integration points of the elements in the finite element model, and the iterative process is started;

[0051] Define the stress σ input measured in the mapping area. Solve a one-step static equilibrium in the first iteration. Due to the introduction of the stress field, the stress components will be redistributed in the mapping area and the non-mapping area;

[0052] Define σ ouput as the redistributed stress in the mapping area after the static equilibrium step. Judge σ input and σouput Is there a deviation between them? If σ input and σ ouput have a deviation, it is determined that there is a redistributed stress in the mapping area after the static equilibrium step. If σ input and σ ouput have no deviation, it is determined that no stress redistribution occurs;

[0053] Taking the mean square error between the reconstructed residual stress value σ ouput in the mapping area and the measured residual stress value σ input of the component as the fitness function, iteratively calculate the population with the minimum fitness. The mean square error formula:

[0054]

[0055] where n is the number of data points, Y i is the i-th actual observed value, is the i-th predicted value;

[0056] Furthermore, to ensure that σ ouput finally converges to σ input after several iterations, the proportional-integral adjustment method is used to modify the output stress to its target measured value during subsequent iterations. The adjustment equation can be expressed as:

[0057]

[0058] where the superscript i represents the i-th iteration, β is the integral factor, and β takes the value of 1, is the initial input stress tensor, that is, the measured convergence value; after the first iteration, the input stress and output stress can be obtained for the next iteration. Repeat to introduce σ input into the mapping area, while the rest of the entire area remains unchanged until and σ input the tolerance between them is less than a predetermined value, so that the full-field stress distribution can be obtained. The implementation procedure of performing finite element analysis using finite element software, using a script written in a parametric design language to post-process the results of the previous iteration, and writing an input file for the next iteration. During the entire iteration process, the mesh always remains unchanged, and the stress components are assigned to the Gauss integration points of the elements.

[0059] Furthermore, as Figure 4As shown in the figure, a residual stress area reconstruction system, the overall framework of the system includes: a user terminal, an interactive interface, and a main program. The user terminal is the input port for residual stress test data at the laser forging repair site. The input data includes: network node data, simulation file models, and measured stress data. The interactive interface is the data transmission interface between the user terminal and the main program. The main program includes an interactive system main program and a stress reconstruction main program. The main program can execute a computer program to implement a residual stress area reconstruction method described in any of the above methods.

[0060] The stress reconstruction main program includes the operation of multiple modules, including: an evaluation fitness module, a main program operation module, a running simulation module, a creating or loading neural network model module, a data processing class module, and a neural network and standardization module;

[0061] The functions of the modules are as follows:

[0062] Evaluation fitness module: Calculate the fitness of individuals in the genetic algorithm. Specifically, calculate the relative error here, and predict the error through a neural network;

[0063] Main program operation module: The main program executes the loop of the genetic algorithm, including initializing the population, evaluating the fitness of individuals, performing selection, crossover, and mutation operations, and checking whether the termination conditions are met (such as reaching the maximum number of generations or the error being lower than the threshold);

[0064] Running simulation module: Call the simulation software and output the results to a specific file for further error calculation;

[0065] Creating or loading neural network model module: Create a new neural network model or load an existing model according to the settings to provide an optimization direction for the genetic algorithm;

[0066] Data processing class module: Responsible for calculating the stress distribution, which is the mathematical model used by the main program when calculating stress, updating the stress data, and providing an evaluation basis for the genetic algorithm;

[0067] Neural network and standardization module: Manage the creation and loading of neural networks, assist in error prediction and model training, provide standardization support for the training data of neural networks, and ensure the stability of the data;

[0068] Preferably, the software interface is as Figure 5 Figure 6 shown, and includes a console window and a parameter definition window. In the parameter definition window, the start and exit of the reconstruction can be controlled, and the initial value of the inherent strain parameter, the number of times the genetic algorithm runs, and the average relative error threshold can be given manually. Information such as the iterative operation process and the number of iterations with the minimum error will be displayed in the console window;

[0069] The specific operation process includes:

[0070] 1. Data extraction: Establish a connection between the residual stress testing equipment and the computer using a USB device. The USB driver needs to be installed before the connection. After the residual stress testing equipment is connected to the computer, the data in the instrument can be extracted. Before running the reconstruction, the data of all limited measurement points need to be extracted and stored. It is necessary to read three files: ('test2.inp'), ('WALL stress data.xlsx'), and ('true S33.txt'). 'test2.inp' is the input file of the finite element model corresponding to the actual stress measurement object, 'WALL stress data.xlsx' is the mesh node data of this finite element model, and 'true S33.txt' contains the data of all limited measurement points. The three files are placed in the directory with the same name as the "Residual Stress Region Reconstruction System Software".

[0071] 2. Reconstruction operation: Double-click to open the software and input the set parameters.

[0072] Input the assumed given initial values for the unknown parameters, and set the inherent strain parameters (range): Cx(0, 4); Dx(0, 10); Cy(0, 4); Dy(0, 10); aL(100, 400); aT(-100, 0).

Claims

1. A residual stress region reconstruction method, characterized in that: The method comprises the following steps: Read the residual stress test data on site; The initial unknown parameters were manually input, the average relative error threshold was calculated by genetic algorithm, and the number of genetic algorithm runs was recorded; The regional residual stress reconstruction is realized through iterative calculation based on the residual stress test data on site, and finally the finite element simulation file of each iteration, the average residual stress value of the corresponding point and the number of iterations at which the minimum error occurs are output and output in the form of charts.

2. A residual stress area reconstruction method according to claim 1, characterized in that: The main steps of the data reading residual stress test data are: obtaining the residual stress distribution of the repaired component at a certain key cross-section position or multiple positions by X-ray diffraction method; A three-dimensional model with the same geometric dimensions as the component is established in the finite element software, a finite element mesh is generated, the mesh node coordinate information is extracted and sent to the scientific computing software for numerical fitting processing, and the parameters of the area where the inherent strain is located are initialized.

3. A residual stress area reconstruction method according to claim 2, characterized in that: The manually input initial unknown parameter is the initialized inherent strain, and the finite element software is called by using the optimization algorithm based on artificial intelligence genetic algorithm and fuzzy finite element theory to perform batch calculation of residual stress reconstructed according to the inherent strain. The mean square error MSE between the residual stress value of the measured section point and the finite element calculated value is used as the fitness function, and the population with the minimum fitness is obtained through iterative calculation, that is, the optimal parameters of the area where the inherent strain is located are obtained; by applying the calculated inherent strain distribution in the finite element model, static analysis is performed, so as to obtain the full-field residual stress distribution with the minimum error with the residual stress on the actual measured section.

4. A residual stress area reconstruction method according to claim 1, characterized in that: The stress values ​​of the sample points in the region containing all the inherent strains are measured from the experiment, and the region is defined as the mapping region, while the region not containing the inherent strains is defined as the non-mapping region. The stress components at each position in the mapping region in the reconstructed finite element model are obtained by multidimensional least squares fitting. The stress components are assigned to the Gaussian integration points of the units in the finite element model, and the iteration process is started; Define the stress σ measured in the mapping area input , in the first iteration, the static equilibrium is solved. Due to the introduction of the stress field, the stress components will be redistributed in the mapping area and the non-mapping area; Define σ ouput To redistribute stress in the mapping region after the static equilibrium step, determine σ input and σ ouput Is there a deviation between input and σ ouput If there is a deviation, it is judged as the redistributed stress in the mapping area after the static equilibrium step. input and σ ouput If there is no deviation, it is judged that no stress redistribution occurs; Reconstruct the residual stress value σ in the mapping area ouput The residual stress value σ input The mean square error is the fitness function, and the population with the minimum fitness is iteratively calculated. The mean square error formula is: Where n is the number of data points, Y i is the actual observation value of the ith is the i-th predicted value.

5. A residual stress area reconstruction method according to claim 4, characterized in that: To ensure that after several iterations σ ouput Finally it converges to σ input , the output stress is modified to its target measured value during subsequent iterations using the proportional-integral adjustment method. The adjustment equation can be expressed as: The superscript i indicates the i-th iteration, β is the integration factor, and β takes the value of 1. is the initial input stress tensor, i.e., the measured convergence value; after the first iteration, the input stress and output stress can be obtained for the next iteration, and σ is repeatedly input The mapping area is introduced, while the rest of the entire region remains unchanged until and σ input The tolerance between them is less than the predetermined value, so that the stress distribution of the whole field can be obtained. ,The implementation procedure of finite element analysis is performed using finite element software. The script written in parametric design language is used to post-process the results of the previous iteration and write the input file for the next iteration. During the entire iteration process, the mesh remains unchanged and the stress components are assigned to the Gaussian integration points of the unit.

6. A residual stress area reconstruction system, characterized in that: The overall framework of the system includes: a user end, an interactive interface and a main program. The user end is an entry for inputting residual stress test data of the laser forging repair site. The input data includes: network node data, simulation file model and measured stress data. The interactive interface is a data transmission interface between the user end and the main program. The main program includes an interactive system main program and a stress reconstruction main program. The main program can execute a computer program to implement a residual stress area reconstruction method described in any one of claims 1-5.

Images