A method for predicting post-weld performance of a steel material, a storage medium, and an electronic device

Abstract

The present application relates to the technical field of intelligent prediction and calculation simulation of steel materials, and particularly relates to a steel material post-weld performance prediction method, a storage medium and an electronic device. The method comprises the following steps: inputting the chemical composition features, metallurgical physical feature parameters, CCT phase change features, welding thermal cycle features, diffusion state variables, microstructure-performance correlation features and cracking index into a physical information artificial intelligence model to establish a performance prediction model based on phase change dynamics, element diffusion dynamics and microstructure evolution law constraints; training the performance prediction model by using a composite physical constraint loss function, and predicting the microstructure, hardness, yield strength, tensile strength and comprehensive cracking risk index of the heat-affected zone of the steel material after welding by using the trained performance prediction model. CRI The present application can realize accurate and reliable prediction of the microstructure and mechanical properties of steel materials after welding.

Description

Technical Field

[0001] This invention relates to the field of intelligent prediction and computational simulation technology for steel materials, and in particular to a method for predicting the post-weld performance of steel materials, a storage medium, and an electronic device. Background Technology

[0002] Steel materials are widely used in engineering structures. With the development of high-strength steel, microalloyed steel, and easy-to-weld steel, higher requirements are placed on the accurate prediction of the post-weld performance of steel materials. During the welding process, steel materials undergo rapid heating and cooling. The weld heat-affected zone is prone to complex metallurgical behaviors such as phase transformation, grain changes, and diffusion of alloying elements, which in turn affect the strength, toughness, and service performance of the welded joint.

[0003] Currently, the post-weld performance evaluation of steel materials mainly relies on experimental testing methods, which suffer from problems such as long cycles, high costs, and difficulty in meeting the needs of rapid material development. In recent years, artificial intelligence technology has been applied to material performance prediction, achieving rapid prediction by establishing the relationship between composition, process, and performance. However, existing artificial intelligence methods mainly rely on data-driven approaches and lack consideration for phase transformation kinetics, element diffusion behavior, and microstructure evolution during the welding process, resulting in insufficient physical interpretability of the models and limited prediction accuracy and reliability under complex welding conditions.

[0004] Traditional computational simulation methods can describe material phase transformation and diffusion processes, but they struggle to directly correlate microstructure with macroscopic properties. Therefore, it is necessary to combine the continuous cooling transformation (CCT) behavior of steel materials, welding thermal cycling processes, element diffusion kinetics, and artificial intelligence models to construct a performance prediction method with physical constraints.

[0005] Based on this, the present invention proposes a method, storage medium and electronic device for predicting the post-weld performance of steel materials based on the fusion of phase transition-diffusion dual physical constraints and artificial intelligence. By introducing CCT phase transition characteristics and diffusion state variables, the accuracy and reliability of predicting the post-weld microstructure and mechanical properties of steel materials are improved. Summary of the Invention

[0006] The purpose of this invention is to provide a method for predicting the post-weld performance of steel materials, a storage medium, and an electronic device. The specific technical solution is as follows: In a first aspect, the present invention provides a method for predicting the post-weld properties of steel materials, comprising: Step S1: Input the chemical composition characteristics, metallurgical physical characteristic parameters, CCT phase transformation characteristics, welding thermal cycle characteristics, diffusion state variables, microstructure-performance correlation characteristics, and cracking index of steel materials into the physical information artificial intelligence model to establish a performance prediction model based on the constraints of phase transformation dynamics, element diffusion dynamics, and microstructure evolution laws. Step S2: Train the performance prediction model using a composite physical constraint loss function. The trained performance prediction model then predicts the microstructure, hardness, yield strength, tensile strength, and comprehensive cracking risk index of the heat-affected zone of steel materials after welding. .

[0007] Optionally, the metallurgical physical characteristic parameters include carbon equivalent, hardenability parameters, austenite stability parameters, and phase transformation driving force parameters; The microstructure-property correlation features include grain size, phase ratio, precipitated phase state, solid solution strengthening parameters, grain refinement strengthening parameters, and phase transformation strengthening parameters. The cracking index includes the hardening cracking index. Martensitic cracking index Carbon equivalent cracking index Cooling cracking index Diffusion hydrogen cracking index and stress cracking index .

[0008] Optionally, the CCT phase transformation characteristics include the phase transformation start temperature of ferrite, pearlite, bainite, and martensite structures of steel materials under different cooling conditions after welding, as output by the CCT curve prediction model. The method for obtaining the CCT curve prediction model includes: collecting and constructing a steel material database; obtaining chemical composition and CCT phase transition data of different steel materials from the steel material database, establishing a mapping relationship between the two, calculating and obtaining metallurgical physical characteristic parameters based on the chemical composition data, and establishing the CCT curve prediction model using a multilayer perceptron neural network through the introduced cooling rate. The steel material database includes chemical composition data of steel materials, CCT test data of continuous cooling transformation, microstructure data at different cooling rates, hardness data at different cooling rates, welding process parameter data, welding thermal cycle data, microstructure data of post-weld heat-affected zone, post-weld mechanical property data, and welding cracking test data.

[0009] Optionally, the welding thermal cycle characteristics include peak temperature during the welding process, high-temperature residence time, cooling rate, and Time; among which, It refers to the time it takes for the temperature of the weld heat-affected zone to drop from 800℃ to 500℃ during the cooling process.

[0010] Optionally, the diffusion state variables include element concentration distribution, element diffusion distance, and element concentration gradient.

[0011] Optionally, the composite physical constraint loss function includes phase transition kinetic constraints, element diffusion constraints, microstructure evolution constraints, strengthening mechanism constraints, and cracking risk constraints; the composite physical constraint loss function is expressed by Equation 1): Formula 1); In Equation 1), represents the phase transition kinetic constraint and satisfies . The phase transformation initiation temperature of each component shows a monotonically decreasing trend with increasing cooling rate, which conforms to the basic metallurgical law of continuous cooling transformation. , , ,as well as All are greater than or equal to zero; The volume fraction of ferrite structure; This represents the volume fraction of pearlite in the tissue. This represents the volume fraction of bainite. This represents the volume fraction of martensite. yes Weighting coefficients; Represents the element diffusion constraint and satisfies ; Indicates the first Concentration of elements With welding thermal cycle time t The rate of change; Indicates the first The spatial gradient of the migration flux of microalloying elements; the diffusion distance of microalloying elements does not exceed the spatial range of the heat-affected zone, which is consistent with the diffusion characteristics of short-time thermal cycling in welding; yes The weighting coefficients are monotonically increased with increasing temperature, which conforms to the Arrhenius diffusion law; This indicates the constraints on microstructure evolution, satisfying the following: coarse-grained heat-affected zone > fine-grained heat-affected zone > spatial distribution of grain size in the steel matrix; with increasing cooling rate, the volume fractions of bainite and martensite monotonically increase, while the volume fractions of ferrite and pearlite monotonically decrease; with increasing microalloying element content, austenite grain growth is suppressed, and grain size tends to decrease, conforming to the basic law of grain boundary pinning; yes Weighting coefficients; This indicates a strengthening mechanism constraint, satisfying the following: both hardness and strength monotonically increase with increasing grain refinement, solid solution element content, precipitate ratio, and martensite volume fraction; the total hardness satisfies the superposition rule. The total strengthening amount is a linear superposition of grain refinement strengthening, solid solution strengthening, precipitation strengthening, and phase transformation strengthening of steel materials, which conforms to the mixed law of steel material strengthening. This indicates the predicted Vickers hardness result; Indicates the matrix hardness of steel materials; Hardness indicates the contribution of grain refinement and strengthening to steel materials; Hardness indicates the contribution of solid solution strengthening to steel materials; Hardness indicates the contribution of precipitation strengthening to steel materials; Hardness indicates the contribution of phase transformation strengthening to steel materials; yes Weighting coefficients; This represents the cracking risk constraint and satisfies the following factors: maximum hardness of the steel material, volume fraction of martensite, carbon equivalent, diffusible hydrogen content, and residual tensile stress, all of which affect the comprehensive cracking risk index. It has a positive impact, and the cooldown time has an effect on... It has a negative impact; for The weighting coefficients.

[0012] Optionally, the strengthening mechanism parameters used in the strengthening mechanism constraint include grain refinement strengthening parameters, solid solution strengthening parameters, precipitation strengthening parameters, and phase transformation strengthening parameters.

[0013] Optionally, the parameters for training the performance prediction model using the composite physical constraint loss function include: the optimization algorithm is adam, the learning rate is 0.00001~0.01, the batch size is 16~256, the number of training iterations is 100~500, and the activation function is ReLU.

[0014] In a second aspect, the present invention provides a computer storage medium storing computer program instructions, which, when executed by a processor, implement the method for predicting the post-weld performance of steel materials.

[0015] In a third aspect, the present invention provides an electronic device comprising: at least one processor, at least one memory, and computer program instructions stored in the memory, wherein the computer program instructions, when executed by the processor, implement the method for predicting the post-weld performance of steel materials.

[0016] The application of the technical solution of the present invention has at least the following beneficial effects: (1) The present invention provides a method for predicting the post-weld performance of steel materials. In step S1, the chemical composition characteristics, metallurgical physical characteristic parameters, CCT phase transformation characteristics, welding thermal cycle characteristics, diffusion state variables, and microstructure-property correlation characteristics of steel materials are input into a physical information artificial intelligence model to establish a performance prediction model based on phase transformation dynamics, element diffusion dynamics, and microstructure evolution laws. In step S2, the performance prediction model is trained using a composite physical constraint loss function. The trained performance prediction model is used to predict the microstructure, hardness, yield strength, tensile strength, and comprehensive cracking risk index of the heat-affected zone of steel materials after welding. This invention enables accurate and reliable prediction of the post-weld microstructure and mechanical properties of steel materials. It is applicable to the evaluation and prediction of the weldability of high-strength steel, microalloyed steel, and free-welding steel.

[0017] (2) The computer storage medium provided by the present invention is used to store computer program instructions. When the computer program instructions are executed by the processor, the method for predicting the post-weld performance of steel materials is implemented, thereby realizing the accurate and reliable prediction of the post-weld microstructure and mechanical properties of steel materials.

[0018] (3) The electronic device provided by the present invention includes at least one processor, at least one memory and computer program instructions stored in the memory. When the computer program instructions are executed by the processor, the method for predicting the post-weld performance of steel materials is implemented, thereby achieving accurate and reliable prediction of the post-weld microstructure and mechanical properties of steel materials.

[0019] In addition to the objectives, features, and advantages described above, the present invention has other objectives, features, and advantages. The invention will now be described in further detail with reference to the figures. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a method for predicting the post-weld performance of steel materials in Example 1.

[0022] Figure 2 This is a predicted CCT curve of the weld joint of HRB600E steel in Example 1.

[0023] Figure 3This is a predicted diffusion distribution diagram of element C at the weld joint of HRB600E steel in Example 1.

[0024] Figure 4 This is a predicted diffusion distribution diagram of Si element in the weld joint of HRB600E steel in Example 1.

[0025] Figure 5 This is a predicted diffusion distribution diagram of Mn element in the weld joint of HRB600E steel in Example 1.

[0026] Figure 6 This is a predicted diffusion distribution diagram of V element at the weld joint of HRB600E steel in Example 1.

[0027] exist Figure 2 In the figure, the horizontal axis Time (sec) represents time; the vertical axis Temperature represents temperature; the Variant in the legend represents different phase transformation microstructures, used to characterize the microstructure transformation behavior of HRB600E steel weld joints during continuous cooling; the legend also includes... F , P , B , and M The terms ferrite, pearlite, bainite, austenite, and martensite are represented respectively. exist Figures 3-6 In the figure, the horizontal axis Distance (mm) represents the diffusion distance; the vertical axis W (C) (%) represents the mass fraction of C element; W (C) in the legend represents the mass fraction of C element; the vertical axis W (Si) (%) represents the mass fraction of Si element; W (Si) in the legend represents the mass fraction of Si element; the vertical axis W (Mn) (%) represents the mass fraction of Mn element; W (Mn) in the legend represents the mass fraction of Mn element; the vertical axis W (V) (%) represents the mass fraction of V element; W (V) in the legend represents the mass fraction of V element. Detailed Implementation

[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0029] Example 1: See Figure 1 A method for predicting the post-weld performance of steel materials, comprising: Step S1: Input the chemical composition characteristics, metallurgical physical characteristic parameters, CCT phase transformation characteristics, welding thermal cycle characteristics, diffusion state variables, microstructure-performance correlation characteristics, and cracking index of steel materials (specifically, HRB600E high-strength easy-to-weld rebar produced by Hunan Iron and Steel Group as the research object) into the physical information artificial intelligence model, and establish a performance prediction model based on the constraints of phase transformation dynamics, element diffusion dynamics, and microstructure evolution law. Step S2: Train the performance prediction model using a composite physical constraint loss function. The trained performance prediction model then predicts the microstructure, hardness, yield strength, tensile strength, and comprehensive cracking risk index of the heat-affected zone of steel materials after welding. .

[0030] The metallurgical physical characteristic parameters include carbon equivalent, hardenability parameters, austenite stability parameters, and phase transformation driving force parameters.

[0031] The microstructure-property correlation features include grain size, phase ratio, precipitated phase state, solid solution strengthening parameters, grain refinement strengthening parameters, and phase transformation strengthening parameters. The cracking index includes the hardening cracking index. Martensitic cracking index Carbon equivalent cracking index Cooling cracking index Diffusion hydrogen cracking index and stress cracking index .

[0032] The CCT phase transformation characteristics include the phase transformation initiation temperatures of ferrite (approximately 720°C), pearlite (approximately 680°C), bainite (approximately 520°C), and martensite (approximately 350°C) of steel under different cooling conditions after welding, as output by the CCT curve prediction model. The volume fraction of ferrite (denoted as...) ), volume fraction of pearlite (denoted as ), volume fraction of bainite structure (denoted as ), and the volume fraction of martensite (denoted as ...). );in, , , , ,as well as All are greater than or equal to zero; The method for obtaining the CCT curve prediction model includes: collecting data from the NIST database to construct a steel material database; obtaining chemical composition and CCT phase transition data of different steel materials from the steel material database, establishing a mapping relationship between the two, calculating metallurgical physical characteristic parameters based on the chemical composition data, and establishing the CCT curve prediction model using a multilayer perceptron neural network through the introduced cooling rate. The parameters used to establish the CCT curve prediction model using the multilayer perceptron neural network include: the optimization algorithm is adam, the learning rate is 0.001, the batch size is 64, the number of training iterations is 200, and the activation function is ReLU. The input layer of the multilayer perceptron neural network has 13 nodes, specifically: 8 chemical composition parameters (see Table 1 for details), 4 metallurgical physical characteristic parameters (carbon equivalent, hardenability parameter, austenite stability parameter, and phase transformation driving force parameter) and 1 cooling rate parameter; the hidden layer includes a first hidden layer (128 neurons) and a second hidden layer (64 neurons); the output layer of the multilayer perceptron neural network is used to output the CCT phase transformation characteristics.

[0033] The steel material database includes chemical composition data of steel materials, CCT test data of continuous cooling transformation, microstructure data at different cooling rates, hardness data at different cooling rates, welding process parameter data (see Table 2 for details), welding thermal cycle data, microstructure data of the heat-affected zone after welding, mechanical property data after welding, and welding cracking test data; the mechanical property data after welding includes hardness, yield strength, and tensile strength data of the heat-affected zone after welding.

[0034] The chemical composition data of HRB600E steel material used in this embodiment 1 includes the mass content of C, Si, Mn, P, S, Cr, Nb and V elements (the mass content of each element is statistically calculated by mass fraction, and the statistical results are shown in Table 1). Based on the chemical composition data, the carbon equivalent, hardenability parameter, austenite stability parameter and phase transformation driving force parameter are further calculated, that is, the metallurgical physical characteristic parameters.

[0035] Table 1 Chemical composition data of HRB600E

[0036] In Table 1, C is mainly used to characterize the hardening tendency of steel materials; Mn and Cr are used to characterize the stability and hardenability of austenite; Nb and V are used to characterize the effects of microalloying, precipitation strengthening and grain refinement.

[0037] HRB600E steel material achieves high strength through C-Mn solid solution strengthening and Nb and V microalloying. Its weldability is mainly affected by phase transformation behavior, element diffusion and precipitation strengthening state under welding thermal cycling.

[0038] Table 2 Welding process parameters for HRB600E

[0039] For HRB600E steel material, multiple continuous cooling rate conditions are set, and different cooling paths are applied to the CCT curve predicted by the CCT curve prediction model.

[0040] At any cooling rate Down( This indicates the cooling rate sequence number, and its value range is... =1, 2, ..., N', where N' is the number of cooling rates set in the CCT curve prediction model. The phase transition type and degree are determined based on the intersection of the cooling path and the CCT phase transition region, and the results are obtained. (indicating the first) (volume fraction of ferrite at each cooling rate) (indicating the first) (volume fraction of pearlite at each cooling rate) (indicating the first) (volume fraction of bainite at each cooling rate) and (indicating the first) (Volume fraction of martensite at different cooling rates); mapping relationship between different chemical compositions, different cooling rates, and different microstructure ratios. Establish an organizational prediction model. As the cooling rate increases, the cooling path of steel materials gradually avoids the high-temperature ferrite and pearlite transformation regions and enters the bainite and martensite transformation regions, thereby increasing the proportion of bainite and martensite.

[0041] After obtaining the microstructure proportions at different cooling rates, a microstructure-hardness correlation model was established using Equation 2).

[0042] Equation 2); In equation 2), Indicates the predicted hardness of steel materials; Indicates the original austenite grain size of steel materials; Indicates the content of solid solution elements; Indicates a precipitation-enhanced state; Indicates the cooling rate.

[0043] The hardness prediction features output by the microstructure-hardness correlation model include ferrite ratio, pearlite ratio, bainite ratio, martensite ratio, original austenite grain size, solid solution element content, microalloying element precipitation state, and cooling rate.

[0044] To improve the physical interpretability of the tissue-hardness correlation model, equation 2) can be further decomposed into equation 3). Equation 3); In equation 3), This indicates the predicted Vickers hardness result; Indicates the matrix hardness of steel materials; Hardness indicates the contribution of grain refinement and strengthening to steel materials; Hardness indicates the contribution of solid solution strengthening to steel materials; Hardness indicates the contribution of precipitation strengthening to steel materials; Hardness represents the contribution of phase transformation strengthening to steel materials.

[0045] The microstructure-hardness correlation model shown in Equations 2 and 3) is the basis for subsequent hardness prediction and cracking risk assessment in the weld heat-affected zone.

[0046] The welding thermal cycle characteristics include peak temperature during the welding process, high-temperature dwell time, cooling rate, and... Time; among which, It refers to the time it takes for the temperature of the weld heat-affected zone to drop from 800℃ to 500℃ during the cooling process.

[0047] The welding thermal cycle characteristics were loaded into the CCT curve prediction model of HRB600E steel, and the phase transformation behavior was determined based on the peak temperature and cooling path at different locations in the welding heat-affected zone.

[0048] The heat-affected zone is divided into: (1) The peak temperature of the coarse-grained heat-affected zone (CGHAZ) is higher than that of the coarse-grained heat-affected zone (CGHAZ). ( This indicates the critical temperature at which ferrite completely transforms into austenite when hypoeutectoid steel is heated (i.e., the critical temperature for complete austenitization) and falls within a relatively high austenitizing temperature range. Complete austenitization occurs in this region, accompanied by the growth of the original austenite grains.

[0049] (2) The peak temperature of the fine-grained heat-affected zone FGHAZ is higher than that of the fine-grained heat-affected zone FGHAZ. However, this temperature is below the coarsening temperature range of CGHAZ. Complete austenitization occurs in this region, but the grain size is relatively small.

[0050] (3) Critical heat-affected zone ICHAZ, its peak temperature In: Partial austenitization occurs, eventually forming a multiphase structure. This indicates the critical temperature at which pearlite begins to transform into austenite when steel is heated (i.e., the critical temperature for eutectoid transformation).

[0051] Based on the heat-affected zone zoning method described above, and combined with the coupling calculation of the welding cooling path and CCT curve, the proportions of ferrite, pearlite, bainite, and martensite in CGHAZ, FGHAZ, and ICHAZ were obtained.

[0052] The simulation of element diffusion kinetics induced by welding thermal cycle is performed, and the specific steps are as follows: A diffusion model of the weld zone of HRB600E steel was established using the Thermo-Calc DICTRA module. The calculation system was an Fe-based multi-element alloy system, utilizing iron-based thermodynamic and diffusion databases. The simulation regions included the weld zone, fusion zone, CGHAZ, and FGHAZ. Welding thermal cycles were used as the basis for the model. As a temperature boundary condition, a one-dimensional diffusion model with a total length of 2 mm is established in the spatial direction, covering a 1 mm region on each side of the fusion line, i.e. The initial conditions used the actual chemical composition of the HRB600E base material, and the weld side used the ER60-G deposited metal composition as the concentration boundary condition. ;in, Indicates the first The concentration of a certain element, when When representing C, Mn, Nb, and V elements, the corresponding initial concentrations are based on the actual test results of the material. x Indicates spatial location; t This indicates the welding thermal cycle time.

[0053] DICTRA solves for the diffusion processes of multiple elements based on temperature-time thermal cycles, thermodynamics, and diffusion databases. Therefore, we obtain the spatial concentration distribution of elements, i.e., ,in, This indicates the end time of the welding thermal cycle (i.e., the termination time of the welding thermal cycle simulation); the change in element concentration, i.e. ; elemental concentration gradient, i.e. ; element diffusion distance, i.e. This represents the spatial migration distance of an element from its peak concentration at the fusion line to ±10% of the initial concentration in the steel base material. ,in, The reference position for the fusion line; element migration flux, i.e. ; Cumulative migration amount of elements, that is, .

[0054] Therefore, DICTRA simulation results are no longer simply represented as a single component of a certain region, but rather form a spatial location... x -time t - Element concentration Three-dimensional state data are used to extract physical state variables such as elemental concentration distribution, diffusion distance, concentration gradient, and migration amount.

[0055] The diffusion state variables include element concentration distribution, element diffusion distance, and element concentration gradient.

[0056] The steps for establishing a performance prediction model based on phase transition kinetics, element diffusion kinetics, and microstructure evolution constraints include: The CCT phase transition characteristics obtained from the CCT curve prediction model are correlated with the element diffusion state variables obtained from DICTRA to form the physical state variables of the weld heat-affected zone: ; in, This indicates a change in element concentration.

[0057] Different regions , , , as well as , , , Input the tissue-hardness correlation model to obtain , , And determine the maximum hardness of the heat-affected zone: ; in, Indicates the hardness of the coarse-grained heat-affected zone; Indicates the hardness of the heat-affected zone of fine grains; This indicates the hardness of the critical heat-affected zone.

[0058] Based on the above-mentioned tissue structure, hardness, and diffusion state, a welding cracking risk assessment model is further established, which includes the following steps: First, regarding the risk of cold cracking in HRB600E welding, the following physical characteristics were selected: HV max , , , , as well as Where CE is the carbon equivalent. This represents the diffusible hydrogen content under actual welding conditions. This refers to the original austenite grain size of the steel material. This refers to the residual tensile stress in the weld heat-affected zone. Secondly, adopt Define hardening crack index ; The critical hardness was determined through actual welding cracking tests on HRB600E. use Define the martensitic cracking index ; The volume fraction of critical martensite determined based on cracking tests; use Define carbon equivalent cracking index ; The critical carbon equivalent obtained from the cracking test; use Define the cooling cracking index ; This refers to the critical cooling time under the corresponding steel materials and welding conditions; where, when the actual When the actual cooling rate is higher than the critical cooling rate, the tendency for hardened structures to form is enhanced. use Define the diffusive hydrogen cracking index ; The critical diffusible hydrogen content determined by the cracking test; use = Define the stress cracking index ; The critical cracking stress was determined through pin tests and residual stress tests. Then, the above cracking index , , , , and Normalization was performed, and a comprehensive cracking risk index was established. That is, the welding cracking risk assessment model; ; And satisfy, ; in, They are respectively , , , , and The weights; Finally, Input the welding cracking risk assessment model and output the comprehensive cracking risk index. , This represents the feature vector for evaluating welding cracking (the set of input features for cracking risk). The comprehensive cracking risk index is used to characterize the cracking sensitivity of steel materials during the welding process and to determine the risk level based on actual cracking risk assessment data.

[0059] The physical information artificial intelligence model used in this embodiment 1 is a feedforward neural network model based on a multilayer perceptron, including an input layer, a hidden layer, a feature fusion layer, and an output layer. The input layer receives input features, namely, the chemical composition features of the steel material, metallurgical physical characteristic parameters, CCT phase transformation features, welding thermal cycle features, diffusion state variables, microstructure-property correlation features, and cracking index. The hidden layer performs nonlinear mapping and feature extraction on the input features. The feature fusion layer fuses feature information corresponding to different physical processes. The output layer outputs the microstructure, hardness, yield strength, tensile strength, and comprehensive cracking risk index of the steel material in the post-weld heat-affected zone. The physical information artificial intelligence model is trained using a composite physical constraint loss function, ensuring that the model's prediction results not only meet the requirements for fitting experimental data, but also satisfy constraints on phase transformation kinetics, element diffusion behavior, microstructure evolution, strengthening mechanisms, and cracking risk of steel materials.

[0060] In other embodiments, the physical information artificial intelligence model may also employ other artificial neural network models, including but not limited to BP neural networks, convolutional neural networks, recurrent neural networks, Transformer networks, or other artificial intelligence models capable of nonlinear mapping of multi-source features.

[0061] The composite physical constraint loss function includes experimental data constraints, phase transition kinetics constraints, element diffusion constraints, microstructure evolution constraints, strengthening mechanism constraints, and cracking risk constraints; the composite physical constraint loss function is expressed by Equation 1): Formula 1); In equation 1), Represents the phase transition dynamics constraints and satisfies The phase transformation initiation temperature of each component shows a monotonically decreasing trend with increasing cooling rate, which conforms to the basic metallurgical law of continuous cooling transformation. , , ,as well as All are greater than or equal to zero; The volume fraction of ferrite structure; This represents the volume fraction of pearlite in the tissue. This represents the volume fraction of bainite. This represents the volume fraction of martensite. yes Weighting coefficients; Represents the element diffusion constraint and satisfies ; Indicates the first Concentration of elements With welding thermal cycle time t The rate of change; Indicates the first The spatial gradient of the migration flux of the elements; the diffusion distance of microalloying elements (such as elements V and Nb) does not exceed the spatial range of the heat-affected zone, which is consistent with the diffusion characteristics of short-time thermal cycling in welding; yes The weighting coefficients are monotonically increased with increasing temperature, which conforms to the Arrhenius diffusion law; This indicates the constraints on microstructure evolution, satisfying the following: coarse-grained heat-affected zone > fine-grained heat-affected zone > spatial distribution of grain size in the steel matrix; with increasing cooling rate, the volume fractions of bainite and martensite monotonically increase, while the volume fractions of ferrite and pearlite monotonically decrease; with increasing microalloying element content, austenite grain growth is suppressed, and grain size tends to decrease, conforming to the basic law of grain boundary pinning; yes Weighting coefficients; This indicates a strengthening mechanism constraint, satisfying the following: both hardness and strength monotonically increase with increasing grain refinement, solid solution element content, precipitate ratio, and martensite volume fraction; the total hardness satisfies the superposition rule. The total strengthening amount is a linear superposition of grain refinement strengthening, solid solution strengthening, precipitation strengthening, and phase transformation strengthening of steel materials, which conforms to the mixed law of steel material strengthening. yes Weighting coefficients; This represents the cracking risk constraint and satisfies the following factors: maximum hardness of the steel material, volume fraction of martensite, carbon equivalent, diffusible hydrogen content, and residual tensile stress, all of which affect the comprehensive cracking risk index. It has a positive impact, and the cooldown time has an effect on... It has a negative impact; for The weighting coefficients.

[0062] The strengthening mechanism parameters used in the aforementioned strengthening mechanism constraints include grain refinement strengthening parameters, solid solution strengthening parameters, precipitation strengthening parameters, and phase transformation strengthening parameters.

[0063] The parameters for training the performance prediction model using the composite physical constraint loss function include: the optimization algorithm is adam, the learning rate is 0.001, the batch size is 64, the number of training iterations is 200, and the activation function is ReLU.

[0064] Input the chemical composition and welding process parameters of HRB600E steel into the trained performance prediction model, and output: CGHAZ microstructure, FGHAZ microstructure, ICHAZ microstructure, HAZ hardness, yield strength, tensile strength, and comprehensive cracking risk index; the output results are predicted values, see Table 3.

[0065] This embodiment also includes basic performance tests of steel materials, welding thermal cycle tests, continuous cooling transformation determination, and welding cracking sensitivity tests. These tests provide labeled training data for a multi-source database and verify the accuracy of post-weld microstructure, mechanical properties, and cracking risk prediction results. The specific test methods are as follows: 1) Preparation of test materials and samples; Test material: HRB600E high-strength easy-to-weld threaded steel bars, nominal diameter Φ25 mm, delivered in hot-rolled condition, and its chemical composition is consistent with that in Table 1.

[0066] Welding sample preparation: Cut a 300 mm long steel bar sample, use a V-shaped butt joint with a bevel angle of 60°, a blunt edge of 2 mm, and an assembly gap of 2 mm; clean the bevel and the area within 20 mm on both sides to ensure the surface is clean.

[0067] Performance testing sampling: After natural cooling for more than 48 hours after welding, metallographic and hardness samples were cut along the cross-section of the welded joint; the heat-affected zone was divided according to the peak temperature and microstructure characteristics, and samples were taken from the coarse-grained heat-affected zone (CGHAZ), fine-grained heat-affected zone (FGHAZ), and critical heat-affected zone (ICHAZ).

[0068] 2) Welding process and thermal cycling test; Welding process parameters: Gas metal arc welding (MAG) is used, with a shielding gas mixture of 80% Ar + 20% CO2 at a flow rate of 15 L / min; ER60-G type solid welding wire with a diameter of Φ1.2 mm is used. The welding current is 220 A, the welding voltage is 26 V, the welding speed is 5 mm / s, single-pass welding is performed, and the interpass temperature is controlled below 150℃.

[0069] Welding thermal cycling test: K-type armored thermocouples were pre-embedded at positions 1 mm, 3 mm, and 5 mm from the fusion line. The sampling frequency was 100 Hz, and the temperature-time curves of the welding process were acquired in real time. Peak temperatures were extracted. High-temperature dwell time and cooling rate.

[0070] 3) Microstructure and mechanical property testing experiments; Quantitative analysis of microstructure: After the metallographic samples were progressively polished with sandpaper and mechanically polished with diamond polishing agent, they were etched with 4% nitric acid alcohol solution for 10-15 s. The microstructure morphology of each region of the heat-affected zone was observed using an optical microscope and a field emission scanning electron microscope. The volume fractions of ferrite, pearlite, bainite, and martensite were quantitatively calculated using image analysis software, and the original austenite grain size was determined using the truncation method. Microhardness testing: A digital display microhardness tester was used with a test load of 100 g and a holding time of 10 s. Tests were conducted at 0.2 mm intervals along the fusion line towards the steel base material to obtain the hardness distribution curve of the heat-affected zone. The characteristic hardness values ​​of CGHAZ, FGHAZ, and ICHAZ, as well as the maximum hardness HVmax of the heat-affected zone, were extracted.

[0071] Tensile property test: Rectangular cross-section specimens were processed according to GB / T 228.1-2021 "Metallic materials - Tensile testing - Part 1: Test method at room temperature" with a gauge length of 50 mm; room temperature tensile test was performed on a universal testing machine with a loading rate of 1 mm / min. The yield strength and tensile strength of the welded joint were determined. Three parallel specimens were set up for each group and the average value was taken.

[0072] 4) Verification of the comprehensive cracking risk index; This experiment is used to calibrate the critical parameters of each component and the threshold of the comprehensive cracking risk index.

[0073] The Y-groove butt joint crack test (Xiaotieyan Test) was conducted according to GB / T 28790-2012 "Test Method for Cold Cracking in Welding - Y-groove Butt Joint Crack Test". Y-groove restraint crack specimens were prepared using the same process parameters as the actual welding. After welding and placing at room temperature for 48 hours, the crack-sensitive characteristic parameters of the specimens were obtained through magnetic particle inspection and cross-sectional observation, including surface crack characteristics, cross-sectional crack characteristics, and root crack characteristics. Three parallel specimens were set up for each process, and the obtained crack-sensitive characteristic parameters were used to determine the comprehensive cracking risk index. Verification; Pin test: The critical fracture stress for cold cracking in the heat-affected zone of HRB600E steel was determined using a pin tester; by adjusting the cooling rate and diffusible hydrogen content, the critical cracking conditions under different working conditions were obtained, and the critical hardness was calibrated. Volume fraction of critical martensite and critical cooling time .

[0074] Diffusible hydrogen content test: The diffusible hydrogen content of the weld metal was determined according to the mercury method in GB / T 3965-2012. The test temperature was 45 ℃ and the collection time was 72 h. The measured value of diffusible hydrogen under the current welding process was obtained and used as the input parameter for the cracking risk index calculation.

[0075] 5) Continuous Cooling Transition (CCT) Curve Measurement Test; The static CCT curve of the test steel was determined using a fully automated phase transformation analyzer to calibrate the CCT prediction model. The sample size was Φ4 mm × 10 mm. It was heated to 1200 ℃ at 10 ℃ / s and held for 5 s to achieve complete austenitization. Seven cooling rates were set: 0.1 ℃ / s, 0.5 ℃ / s, 1 ℃ / s, 5 ℃ / s, 10 ℃ / s, 20 ℃ / s, and 50 ℃ / s. The samples were continuously cooled to room temperature. The thermal expansion curves were collected. Combined with the microstructure observation and hardness test after cooling, the phase transformation start temperature, phase transformation end temperature, and microstructure transformation amount at each cooling rate were determined. The complete CCT curve of the test steel was then plotted.

[0076] 6) Element diffusion distribution test verification: An electron probe microanalyzer (EPMA) was used to perform elemental line scan tests on the weld heat-affected zone. The accelerating voltage was 15 kV, the beam current was 20 nA, and the step size was 0.01 mm. The concentration distribution curves of C, Mn, Nb, and V along the fusion line toward the steel material base were measured. The measured elemental distribution was compared with the DICTRA diffusion kinetics simulation results to calibrate the diffusion activation energy and diffusion coefficient parameters, ensuring the accuracy of the calculation of diffusion state variables.

[0077] Through the above experimental methods, all the experimental value data listed in Table 3 of Example 1 can be obtained, which not only provides training and verification samples for the physical information artificial intelligence model, but also realizes the full-process verification of CCT prediction, tissue evolution, diffusion behavior and crack risk prediction results.

[0078] Table 3 Test values, predicted values, and relative errors of HRB600E post-weld performance.

[0079] As shown in Table 3, the relative error between the predicted and experimental values ​​of the post-weld performance of HRB600E high-strength easy-to-weld threaded steel using the method for predicting the post-weld performance of steel materials in Example 1 is less than 3%.

[0080] See Figure 2 In this embodiment 1, the CCT curve prediction model is used to output the CCT curve prediction diagram of the weld point of HRB600E high-strength easy-to-weld threaded steel. Figure 2This embodiment describes the predicted continuous cooling transformation (CCT) curve of the weld joint of HRB600E high-strength easy-to-weld rebar, including the phase transformation initiation and termination temperatures and microstructure transformation patterns under different cooling rates. The CCT curve prediction results are used to extract the phase transformation kinetic characteristics under welding thermal cycling conditions and serve as CCT phase transformation feature information in the physical information artificial intelligence prediction model to improve the accuracy of post-weld microstructure and property prediction.

[0081] See Figure 3 In this embodiment 1, the diffusion kinetics simulation is used to output the predicted diffusion distribution map of element C at the weld point of HRB600E steel. Figure 3 This embodiment describes the predicted diffusion kinetics distribution of elements at the weld joint of HRB600E steel, reflecting the concentration migration law of element C in the weld joint region under welding thermal cycling. This predicted C diffusion result is used to obtain the characteristics of element migration and concentration changes during welding, and serves as the diffusion state variable information in the physical information artificial intelligence prediction model, thereby achieving accurate prediction of post-weld microstructure evolution and property changes.

[0082] See Figure 4 In this embodiment 1, the diffusion kinetics simulation is used to output the predicted Si element diffusion distribution map of the HRB600E steel weld joint. Figure 4 This embodiment illustrates the predicted diffusion kinetics distribution of Si in the weld joint of HRB600E steel, reflecting the concentration migration pattern of Si in the weld joint region under welding thermal cycling. This Si diffusion characteristic is used to characterize the migration behavior of alloying elements during welding thermal cycling, providing element diffusion characteristic input for predicting post-weld microstructure evolution.

[0083] See Figure 5 In this embodiment 1, the diffusion kinetics simulation is used to output the predicted diffusion distribution map of Mn element in the weld joint of HRB600E steel. Figure 5 This embodiment illustrates the predicted diffusion kinetics distribution of Mn element in the weld joint of HRB600E steel, reflecting the concentration migration law of Mn element in the weld joint region under welding thermal cycling. This Mn element diffusion characteristic is used to establish the correlation between element diffusion behavior and microstructure evolution.

[0084] See Figure 6 In this embodiment 1, the diffusion kinetics simulation is used to output the predicted diffusion distribution map of element V at the weld point of HRB600E steel. Figure 6This embodiment illustrates the predicted diffusion kinetics distribution of V element in the weld joint of HRB600E steel, reflecting the concentration migration law of V element in the weld joint region under welding thermal cycling. This V element diffusion characteristic is used to characterize the migration behavior of microalloying elements and provides physical characteristics for predicting post-weld microstructure and properties and evaluating the Comprehensive Cracking Risk Index (CRI).

[0085] Through the aforementioned CCT phase transformation prediction and element diffusion kinetics simulation, the CCT phase transformation characteristics and diffusion state variables during the welding thermal cycle are obtained. These are then combined with the chemical composition characteristics, metallurgical physical parameters, welding thermal cycle characteristics, microstructure-property correlation characteristics, and cracking index of steel materials. The results are then input into a physical information artificial intelligence prediction model to determine the post-weld microstructure and properties of steel materials and their comprehensive cracking risk index. The prediction.

[0086] Example 2: Unlike Example 1, the research object of Example 1 was replaced with high-vanadium HRB600 high-strength steel bars produced by Hunan Iron and Steel Group (see Table 4 for chemical composition). Welding process parameter data used in the steel material database for high-vanadium HRB600 high-strength steel bars are shown in Table 5; additionally, microalloying precipitation characteristic data such as V (C,N) precipitate size, precipitation volume fraction, and precipitation strengthening contribution were supplemented to the steel material database. In the CCT phase transformation characteristics of high-vanadium HRB600 high-strength steel bars, the phase transformation initiation temperature of ferrite is approximately 705℃, that of pearlite is approximately 660℃, that of bainite is approximately 505℃, and that of martensite is approximately 335℃. Using the same post-weld performance prediction and experimental methods as described in Example 1, the predicted and experimental values ​​of high-vanadium HRB600 high-strength steel bars were measured, and the results are shown in Table 6.

[0087] Table 4 Chemical composition data of high vanadium HRB600

[0088] Table 5 Welding process parameters for high vanadium HRB600

[0089] Table 6. Test values, predicted values, and relative errors of post-weld performance of high-vanadium HRB600

[0090] As shown in Table 6, Example 2, targeting high-vanadium HRB600 high-strength steel bars, verified the accurate characterization ability of the post-weld performance prediction method for steel materials described in Example 1 for the precipitation-dissolution behavior of high-vanadium microalloying elements. It achieved full-process prediction of vanadium diffusion kinetics, microstructure evolution, hardness distribution, and cracking risk under welding thermal cycling. Compared with conventional HRB600E steel, Example 2 can accurately capture the precipitation strengthening effect and hardening tendency changes brought about by high vanadium, making it suitable for welding performance evaluation, process optimization, and composition design of high-vanadium high-strength steel bars.

[0091] Example 3: Unlike Example 1, the research object of Example 1 was replaced with high-niobium HRB600 high-strength steel bars produced by Hunan Iron and Steel Group (see Table 7 for chemical composition). Welding process parameter data used in the steel material database for high-niobium HRB600 high-strength steel bars are shown in Table 8; additionally, microalloy grain boundary characteristic data, such as Nb (C,N) precipitate pinning force, grain boundary migration resistance, and precipitate size distribution, were added to the steel material database. In the CCT phase transformation characteristics of high-niobium HRB600 high-strength steel bars, the phase transformation initiation temperature of ferrite is approximately 695℃, that of pearlite is approximately 650℃, that of bainite is approximately 495℃, and that of martensite is approximately 325℃. Using the same post-weld performance prediction and experimental methods as described in Example 1, the predicted and experimental values ​​of high-niobium HRB600 high-strength steel bars were measured, and the results are shown in Table 9.

[0092] Table 7 Chemical composition data of high-niobium HRB600

[0093] Table 8 Welding process parameters for high-niobium HRB600

[0094] Table 9. Test values, predicted values, and relative errors of post-weld performance of high-niobium HRB600

[0095] As shown in Table 9, Example 3, using high-niobium HRB600 high-strength rebar as the research object, verified the post-weld performance prediction method of steel materials described in Example 1. This method effectively controlled the grain boundary pinning effect and precipitation-dissolution behavior of high-niobium microalloying elements on grain growth, microstructure evolution, and cracking risk in the weld heat-affected zone. The method was verified to be applicable to the prediction of Nb-V composite microalloyed high-strength steel bars.

[0096] The above description is only a preferred embodiment of the present invention and does not limit the scope of the present invention. All equivalent structural transformations made under the inventive concept of the present invention using the contents of the present invention specification and drawings, or direct / indirect applications in other related technical fields, are included within the protection scope of the present invention.

Claims

1. A method for predicting the post-weld performance of steel materials, characterized in that, include: Step S1: Input the chemical composition characteristics, metallurgical physical characteristic parameters, CCT phase transformation characteristics, welding thermal cycle characteristics, diffusion state variables, microstructure-performance correlation characteristics, and cracking index of steel materials into the physical information artificial intelligence model to establish a performance prediction model based on the constraints of phase transformation dynamics, element diffusion dynamics, and microstructure evolution laws. Step S2: Train the performance prediction model using a composite physical constraint loss function. The trained performance prediction model then predicts the microstructure, hardness, yield strength, tensile strength, and comprehensive cracking risk index of the heat-affected zone of steel materials after welding. .

2. The method for predicting the post-weld performance of steel materials as described in claim 1, characterized in that, The metallurgical physical characteristic parameters include carbon equivalent, hardenability parameters, austenite stability parameters, and phase transformation driving force parameters. The microstructure-property correlation features include grain size, phase ratio, precipitated phase state, solid solution strengthening parameters, grain refinement strengthening parameters, and phase transformation strengthening parameters. The cracking index includes the hardening cracking index. Martensitic cracking index Carbon equivalent cracking index Cooling cracking index Diffusion hydrogen cracking index and stress cracking index .

3. The method for predicting the post-weld performance of steel materials as described in claim 1, characterized in that, The CCT phase transformation characteristics include the phase transformation start temperature of ferrite, pearlite, bainite, and martensite structures under different cooling conditions after welding, as output by the CCT curve prediction model. The method for obtaining the CCT curve prediction model includes: collecting and constructing a database of steel materials; Chemical composition and CCT phase transition data of different steel materials are obtained from the steel material database, a mapping relationship between the two is established, and metallurgical physical characteristic parameters are calculated based on the chemical composition data. By introducing the cooling rate, a multilayer perceptron neural network is used to establish the CCT curve prediction model. The steel material database includes chemical composition data of steel materials, CCT test data of continuous cooling transformation, microstructure data at different cooling rates, hardness data at different cooling rates, welding process parameter data, welding thermal cycle data, microstructure data of post-weld heat-affected zone, post-weld mechanical property data, and welding cracking test data.

4. The method for predicting the post-weld performance of steel materials as described in claim 1, characterized in that, The welding thermal cycle characteristics include peak temperature during the welding process, high-temperature dwell time, cooling rate, and... Time; among which, It refers to the time it takes for the temperature of the weld heat-affected zone to drop from 800℃ to 500℃ during the cooling process.

5. The method for predicting the post-weld performance of steel materials as described in claim 1, characterized in that, The diffusion state variables include element concentration distribution, element diffusion distance, and element concentration gradient.

6. The method for predicting the post-weld performance of steel materials as described in any one of claims 1 to 5, characterized in that, The composite physical constraint loss function includes phase transformation kinetics constraints, element diffusion constraints, microstructure evolution constraints, strengthening mechanism constraints, and cracking risk constraints; the composite physical constraint loss function is expressed by Equation 1): Formula 1); In equation 1), Represent the phase transition dynamics constraints and satisfy... The phase transformation initiation temperature of each component shows a monotonically decreasing trend with increasing cooling rate, which conforms to the basic metallurgical law of continuous cooling transformation. , , ,as well as All are greater than or equal to zero; The volume fraction of ferrite structure; This represents the volume fraction of pearlite in the tissue. This represents the volume fraction of bainite. This represents the volume fraction of martensite. yes Weighting coefficients; Represents the element diffusion constraint and satisfies ; Indicates the first Concentration of elements With welding thermal cycle time t The rate of change; Indicates the first The spatial gradient of the migration flux of microalloying elements; the diffusion distance of microalloying elements does not exceed the spatial range of the heat-affected zone, which is consistent with the diffusion characteristics of short-time thermal cycling in welding; yes The weighting coefficients are monotonically increased with increasing temperature, which conforms to the Arrhenius diffusion law; This represents the constraints on organizational evolution, satisfying the following: coarse-grained heat-affected zone > fine-grained heat-affected zone > spatial distribution of grain size in the steel material matrix; As the cooling rate increases, the volume fractions of bainite and martensite increase monotonically, while the volume fractions of ferrite and pearlite decrease monotonically. As the content of microalloying elements increases, austenite grain growth is suppressed, and the grain size tends to decrease, which conforms to the basic law of grain boundary pinning. yes Weighting coefficients; This indicates a strengthening mechanism constraint, satisfying the following: both hardness and strength monotonically increase with increasing grain refinement, solid solution element content, precipitate ratio, and martensite volume fraction; the total hardness satisfies the superposition rule. The total strengthening amount is a linear superposition of grain refinement strengthening, solid solution strengthening, precipitation strengthening, and phase transformation strengthening of steel materials, which conforms to the mixed law of steel material strengthening. This indicates the predicted Vickers hardness result; Indicates the matrix hardness of steel materials; Hardness indicates the contribution of grain refinement and strengthening to steel materials; Hardness indicates the contribution of solid solution strengthening to steel materials; Hardness indicates the contribution of precipitation strengthening to steel materials; Hardness indicates the contribution of phase transformation strengthening to steel materials; yes Weighting coefficients; This represents the cracking risk constraint and satisfies the following factors: maximum hardness of the steel material, volume fraction of martensite, carbon equivalent, diffusible hydrogen content, and residual tensile stress, all of which affect the comprehensive cracking risk index. It has a positive effect, and the cooldown time has a positive impact. It has a negative impact; for The weighting coefficients.

7. The method for predicting the post-weld performance of steel materials as described in claim 6, characterized in that, The strengthening mechanism parameters used in the aforementioned strengthening mechanism constraints include grain refinement strengthening parameters, solid solution strengthening parameters, precipitation strengthening parameters, and phase transformation strengthening parameters.

8. The method for predicting the post-weld performance of steel materials as described in claim 7, characterized in that, The parameters for training the performance prediction model using the composite physical constraint loss function include: the optimization algorithm is Adam, the learning rate is 0.00001~0.01, the batch size is 16~256, the number of training iterations is 100~500, and the activation function is ReLU.

9. A computer storage medium, characterized in that, It stores computer program instructions, which, when executed by a processor, implement the method for predicting the post-weld performance of steel materials as described in claim 8.

10. An electronic device, characterized in that, include: The method for predicting the post-weld performance of steel materials as described in claim 8 includes at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method for predicting the post-weld performance of steel materials as described in claim 8.

Images