A method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel

Abstract

The present invention relates to the field of welding wire microstructure and performance. The present invention provides a method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for ship hull steel, comprising: data segmentation; data standardization; selecting a GBDT algorithm to establish a model; after completing the regression model establishment, obtaining the mass percentages of independent variables C, Mn, Si, Cr, Mo, Ni, Cu, and CE, and t8 / 5 as inputs, and outputting the volume fractions of phase components; and establishing a microstructure model based on the volume fractions of the phase components to predict the mechanical properties of the flux-cored welding wire. The method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for ship hull steel described in the present invention can not only accurately predict the volume fractions of the phase components in the microstructure of the flux-cored welding wire, but also accurately predict the mechanical properties of the flux-cored welding wire.

Description

Technical Field

[0001] The present invention relates to the field of welding wire microstructure and performance, and in particular to a method for predicting microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel. Background Art

[0002] Welding is a fundamental forming method for the manufacture of large steel structures. For shipbuilding, the processability and efficiency of the welding materials used are particularly important. Commonly used welding materials include submerged arc welding wire, manual welding rods, solid welding wire, and flux-cored welding wire. Submerged arc welding wire has high production efficiency, but high heat input, prone to workpiece deformation, and limited submerged arc welding positions. Manual welding rods have good adaptability to welding positions, but their welding efficiency is low and they require high technical skills from welding operators. Gas-shielded solid welding wire has high mechanical properties and low diffusible hydrogen, but it places high demands on welding equipment, and welding processability using ordinary welding machines is poor. Flux-cored welding wire, on the other hand, offers a soft arc and excellent processability in all welding positions. Once invented, it was rapidly adopted and applied in the shipbuilding industry.

[0003] Since most domestic flux-cored welding wires are still general-purpose products such as E501-T1, and ordinary civilian ships mostly use general-strength hull structural steel with low inherent crack sensitivity and low low-temperature toughness requirements, more common acid slag flux-cored wires can meet the needs. However, for certain large ocean-going vessels, coast guard vessels, and polar vessels with strict economic and environmental requirements, Ni-Cr-Mo high-strength and high-toughness low-alloy hull structural steels with a strength of 390MPa or above are often used as the main structural material. Due to the increased strength grade, hull weight can be reduced by using a large number of medium and thin plates with specifications of less than 16mm. This requires the welding materials to have low heat input, low crack sensitivity, and high low-temperature toughness. A major difficulty in matching flux-cored welding wire with high-strength steels such as those used in shipbuilding and offshore engineering is to solve the problem of welding material performance. The key means of performance control is to ensure that the deposited metal alloying design of the flux-cored welding wire achieves a good microstructure within the welding heat input range. A large number of studies have shown that the weld with a structure dominated by acicular ferrite AF can achieve the best comprehensive performance in the strength level of 400~600MPa.

[0004] The current tradition of welding material design is still based on trial and error, lacking quantitative design calculation methods. As an important material prerequisite for ships and offshore structures, their performance level will affect service safety. With the development of digital technology, structural-level design based on CFD and CAE simulation tools can greatly shorten the design cycle of new ship types. However, the application of calculation methods in welding material design has certain limitations: from research and development to manufacturing, the correlation mechanism between each process and each link is complex and there are many coupling factors. Compared with CAD / CAE commonly used in equipment design, the technical maturity and application of computational materials science are relatively low. In addition, the comprehensive service performance requirements of ships and offshore structures are high, so there are relatively few related research and applications.

[0005] Establishing a machine learning model for existing data and skipping the complex mechanism analysis of some links is an effective way to solve the high-dimensional nonlinear problem of welding material performance prediction.

[0006] Patent application number CN201810067169.7 in the prior art discloses a method and control device for predicting welding residual stress based on structural transformation. By changing the composition of the weld metal, the weld structure is transformed. The stress trend during welding is detected using strain gauges to record the influence of the structural transformation on the welding stress, thereby controlling the welding wire structure to predict the internal stress distribution and state of the metal material after welding, reduce the residual stress of the material after welding, and optimize the mechanical properties after welding. Although this patent predicts the relationship between the welding structure and the stress distribution and magnitude by comparing the measured welding stress with the metallographic structure of the sample, it cannot predict the phase components in the microstructure of flux-cored wires with different compositions and process parameters, nor can it predict the mechanical properties of flux-cored wires.

[0007] In view of this, the present invention is proposed. Summary of the Invention

[0008] The purpose of the present invention is to propose a method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel, so as to solve the problem in the prior art that the phase components and mechanical properties in the microstructure of flux-cored welding wire with different compositions and process parameters cannot be predicted.

[0009] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0010] A method for predicting the microstructure characteristics and mechanical properties of a flux-cored welding wire for ship hull steel, the method being applicable to predicting the volume fraction of phase components in the microstructure of the flux-cored welding wire and the mechanical properties of the flux-cored welding wire, the phase components comprising acicular ferrite AF, proeutectoid ferrite PF, bainite B, and martensite M;

[0011] The prediction method comprises the following steps:

[0012] S1. Data division: Accumulate the quantitative analysis data of the welding wire structure, and divide the accumulated quantitative analysis data of the welding wire structure into a training set and a validation set. The training set is used to directly build the regression model, and the validation set is used to evaluate the model during the construction process;

[0013] S2. Data standardization: After data partitioning, the training set and validation set are subjected to Z-score standardization respectively;

[0014] S3, model selection: select GBDT algorithm to build the model;

[0015] S4. Model building: Select model parameters for modeling, compare the weighted output results of the model under different parameters through cross-validation, then use the optimal parameters to build a regression model for the training set, and use the validation set to assess the model's capabilities. Depending on the size of the validation set, optimize the model parameters, and then regress the training set and validation set to finally complete the regression model building;

[0016] S5. Microstructure prediction: After the regression model is established, the mass percentages of the independent variables C, Mn, Si, Cr, Mo, Ni, Cu, and CE and t8 / 5 are obtained as inputs, and the volume fractions of the phase components are output;

[0017] S6. Mechanical properties prediction: A microstructure model is established based on the volume fraction of the phase components to predict the mechanical properties of the flux-cored welding wire.

[0018] The method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel described in the present invention can not only accurately predict the volume fraction of phase components in the microstructure of the flux-cored welding wire, but also accurately predict the mechanical properties of the flux-cored welding wire.

[0019] Furthermore, step S6 is specifically as follows: after obtaining the predicted results of the volume fractions of the phase components of the welding wire structure, firstly, a series of microstructure models that conform to the volume fractions of the phase components are generated by computer according to the predicted results, and a two-point statistical function analysis of the microstructure model is performed using software tools, and the microstructure model with the lowest residual is selected to establish a two-dimensional representative volume unit; then, the engineering stress-strain curve of the welding wire structure is calculated by the crystal plasticity method, and finally, the predicted values of the yield strength and tensile strength of the welding wire structure are obtained according to the calculated values of the engineering stress-strain curve.

[0020] Furthermore, in step S1, the accumulated quantitative analysis data of the welding wire structure includes weld metal with a yield strength of 350-1020 MPa.

[0021] Furthermore, in step S1, the chemical composition of the weld metal includes, by mass percentage, w(C)≤0.1%, w(Mn): 0.60-1.5%, w(Si)≤0.4%, w(Cr)≤0.4%, w(Mo)≤0.5%, w(Ni): 0.6-1.5%, w(V)≤0.05%, w(Cu)≤0.5%, w(S)≤0.0025%, w(P)≤0.0025%, and the remainder is Fe and unavoidable impurities.

[0022] Furthermore, in step S1, the accumulated quantitative analysis data of the welding wire structure are the component analysis results, process parameters, and volume fractions of phase components corresponding to the welding wire structure under different welding processes.

[0023] Furthermore, the composition analysis results include at least the mass percentages of C, Mn, Si, Cr, Mo, Ni, Cu and CE.

[0024] Furthermore, the process parameters at least include t8 / 5.

[0025] Furthermore, in step S1, the accumulated welding wire structure quantitative analysis data is preprocessed, and the data is divided using a leave-one-out method or a k-fold cross validation method, where k≥50.

[0026] Furthermore, in step S2, the calculation steps for Z-score standardization are: new data = (original data - mean) / standard deviation.

[0027] Furthermore, in step S4, when the model capability is assessed on the validation set, it is determined whether the root mean square error of the mechanical properties of the samples in the model prediction validation set is significantly higher than the root mean square error of the training set and the validation set, or whether the data-corrected coefficient of variation of the model prediction value and the validation set is significantly lower than the data-corrected coefficient of variation of the training set; if any one of the conditions is met, it means that the original data distribution has changed or the model selection is unreasonable, and the model needs to be rebuilt; if none of the conditions are met, the model parameters only need to be optimized depending on the size of the validation set.

[0028] Compared with the prior art, the method for predicting the microstructure characteristics and mechanical properties of a flux-cored welding wire for hull steel described in the present invention has the following beneficial effects:

[0029] 1) The method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel described in the present invention can accurately predict not only the phase components in the microstructure of flux-cored welding wire with different compositions and process parameters in actual production, but also the volume fractions of phase components such as acicular ferrite AF, proeutectoid ferrite PF, bainite B, and martensite M.

[0030] 2) The method for predicting the microstructure characteristics and mechanical properties of a flux-cored welding wire for hull steel described in the present invention, after obtaining the predicted results of the volume fractions of the phase components of the welding wire structure, first generates a series of microstructure models that conform to the volume fractions of the phase components by computer, and finds the most reasonable microstructure model based on the measured two-point statistical function of the microstructure. Then, the stress-strain curve of the welding wire structure is calculated using the crystal plasticity method. Finally, based on the calculated values of the stress-strain curve, the mechanical properties of the welding wire structure are predicted, thereby realizing the mechanical property regulation of the flux-cored welding wire. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 This is a schematic diagram of the principle of a method for predicting the microstructure characteristics and mechanical properties of a flux-cored welding wire for hull steel according to an embodiment of the present invention;

[0032] Figure 2 A schematic diagram of a model for establishing a method for predicting the microstructure characteristics and mechanical properties of a flux-cored welding wire for hull steel according to an embodiment of the present invention;

[0033] Figure 3 This is a schematic diagram of the evaluation of the model capability of a validation set for a method for predicting the microstructure characteristics and mechanical properties of a flux-cored welding wire for hull steel according to an embodiment of the present invention;

[0034] Figure 4 This is one of the schematic diagrams of stress-strain curves of the welding wire structure in a method for predicting the structural characteristics and mechanical properties of a flux-cored welding wire for hull steel according to an embodiment of the present invention;

[0035] Figure 5 This is a second schematic diagram of stress-strain curves of a welding wire structure for a method for predicting the structural characteristics and mechanical properties of a flux-cored welding wire for hull steel according to an embodiment of the present invention. DETAILED DESCRIPTION

[0036] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other. The descriptions of "first", "second", etc. mentioned in the embodiments of the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in this field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0037] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0038] Currently, there are few machine learning models for material microstructure prediction, especially for the analysis of high-strength steel welds. In addition, there is also little research on the calculation of mechanical properties of heterogeneous homogeneous welds containing AF / PF / B.

[0039] The existing technology has the problem of being unable to predict the phase components and mechanical properties in the microstructure of flux-cored welding wires with different compositions and process parameters.

[0040] To address the aforementioned technical problems, the present invention aims to solve the difficult problem of weld metal microstructure and property prediction. First, given the numerous influencing variables of marine welding material microstructure, quantitative metallographic data is accumulated, followed by data preprocessing and parameter initialization. The objective function (minimizing the mean error in the validation set for the volume fractions of acicular ferrite, proeutectoid ferrite, and bainite phases, while maximizing the correlation coefficient in the training set) is determined, a hyperparameter search range is defined, and model development is completed. Secondly, the mechanical property prediction problem of typical welding materials is addressed. Based on the predicted weld phase volume fractions, a series of microstructure models that conform to the phase volume fractions are generated by computer. The most reasonable microstructure model is identified based on the measured two-point statistical function of the microstructure. The engineering stress-strain curve of the material is calculated using a crystal plasticity method, and the yield strength and tensile strength are determined based on the calculated engineering stress-strain curve values. These methods provide a means for controlling the properties of 460 MPa-grade flux-cored welding wire.

[0041] The purpose of this invention is to control the structure and performance of welding materials, and to develop a machine learning model and strength analysis method suitable for predicting the structural characteristics of hull steel supporting welding wire. Based on the existing stock data, a structural prediction model is given synchronously, and a structural prediction model is proposed to complete the mechanical property prediction, providing a digital tool for the design of welding materials.

[0042] The overall analysis process is as follows Figure 1 , where tissue distribution analysis is to complete the statistical analysis of the two-point distribution function of the pixel points representing different phases in the metallographic photograph through image processing methods.

[0043] The first step is to conduct a quantitative analysis of the structural characteristics. Most of these properties are related to the chemical composition of the deposited metal and the welding process. Therefore, when establishing the model, key features are selected from influencing factors such as w(C), w(Mn), w(Ni), w(Cr), w(Mo), w(V), w(Cu), tempering temperature, heat input, t8 / 5, t8 / 3, etc., and regression is performed using symbols or machine learning. Machine learning methods can handle more complex data relationships and have relatively high accuracy when the mechanism affecting material properties is unclear. The basic process of machine learning modeling is as follows: Figure 2 shown.

[0044] Specifically, the present invention proposes a method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel.

[0045] The prediction method is applicable to predicting the volume fraction of phase components in the microstructure of flux-cored welding wire and the mechanical properties of the flux-cored welding wire, wherein the phase components include acicular ferrite AF, proeutectoid ferrite PF, bainite B and martensite M; and the mechanical properties include yield strength and tensile strength;

[0046] Specifically, the prediction method is suitable for predicting the volume fraction of phase components in the microstructure of flux-cored welding wire with a yield strength of 460~590MPa; the prediction method can realize the prediction of the volume fraction and mechanical properties of phase components in the microstructure of flux-cored welding wire with a yield strength of 460MPa.

[0047] The prediction method comprises the following steps:

[0048] S1. Data division: Accumulate quantitative analysis data of welding wire structure and divide the accumulated quantitative analysis data of welding wire structure into a training set and a validation set. The training set is used to directly build a regression model, and the validation set is used to evaluate the model during the construction process, provide an unbiased estimate for the model, and further provide a reference for model parameter adjustment;

[0049] Since the data involved in this project is small and the computational overhead required to build the regression model is affordable, in order to fully utilize the existing data, specifically, in step S1, the accumulated quantitative analysis data of welding wire structure are preprocessed and the data are divided using the leave-one-out method or k-fold cross-validation method, where k≥50.

[0050] This method can ensure that every piece of data in the dataset is involved in the verification, and the data used in each round of regression is only one piece less than the original data. The evaluation results are stable and the evaluation deviation is small, which is conducive to the performance of the prediction model in unknown data.

[0051] In step S1, the accumulated quantitative analysis data of the welding wire structure includes weld metal with a yield strength of 350-1020 MPa. This setting can improve the generalization ability of the model.

[0052] In fact, the phase components in the microstructure of the flux-cored welding wire with a yield strength of 460 MPa include acicular ferrite AF, proeutectoid ferrite PF, and bainite B. The phase components in the microstructure of the flux-cored welding wire with a higher yield strength include a small amount of martensite M.

[0053] In step S1, as shown in Table 1, the chemical composition of the weld metal includes, by mass percentage, w(C)≤0.1%, w(Mn): 0.60-1.5%, w(Si)≤0.4%, w(Cr)≤0.4%, w(Mo)≤0.5%, w(Ni): 0.6-1.5%, w(V)≤0.05%, w(Cu)≤0.5%, w(S)≤0.0025%, w(P)≤0.0025%, and the remainder is Fe and unavoidable impurities.

[0054] Table 1 Chemical composition requirements of weld metal

[0055]

[0056] In step S1, the accumulated quantitative analysis data of the welding wire structure are the component analysis results, process parameters, and volume fractions of phase components corresponding to the welding wire structure under different welding processes.

[0057] The composition analysis results include at least the mass percentages of C, Mn, Si, Cr, Mo, Ni, Cu and CE, namely w(C), w(Mn), w(Si), w(Cr), w(Mo), w(Ni), w(Cu) and w(CE).

[0058] Where CE is carbon equivalent, and its formula is as follows:

[0059] CE(IIW)= C + Mn / 6 + (Cu + Ni) / 15 + (Cr + Mo + V) / 5.

[0060] The process parameters at least include t8 / 5.

[0061] t8 / 5 is a crucial parameter in the welding thermal cycle, crucial for controlling weld quality and improving weld performance. The t8 / 5 value determines the microstructure and properties of the fusion zone. Excessively fast or slow cooling rates can lead to weld defects. Therefore, properly controlling the t8 / 5 value is crucial for ensuring weld quality. For non-hardenable steel, t8 / 5 is the cooling time from 800°C to 500°C.

[0062] The volume fractions of the phase components include the volume fraction of acicular ferrite AF, the volume fraction of proeutectoid ferrite PF, the volume fraction of bainite B and the volume fraction of martensite M.

[0063] In this application, w(C), w(Mn), w(Si), w(Cr), w(Mo), w(Ni), w(Cu), and w(CE) represent the mass percentages of C, Mn, Si, Cr, Mo, Ni, Cu, and CE, respectively.

[0064] S2. Data standardization: After data partitioning, the training set and validation set are subjected to Z-score standardization respectively;

[0065] In step S2, the calculation steps for Z-score standardization are: new data = (original data - mean) / standard deviation.

[0066] After Z-score normalization of the training and validation sets, the sample values are dimensionlessly distributed within a range with a mean of 0 and a standard deviation of 1. This eliminates the effect of sample size on the weight of the independent variable, prevents information leakage from the validation set, and, to a certain extent, suppresses the interference of data noise and outliers on the model, accelerating the convergence of the regression algorithm.

[0067] S3, model selection: select GBDT algorithm to build the model;

[0068] GBDT is an additive model based on boosting ensemble learning. It uses a feed-forward distribution algorithm for training. Each iteration learns a CART tree to fit the residual between the predictions of the previous t-1 trees and the true values of the training samples. This model has strong interpretability and proven effectiveness in practice. It is often used in the top three big data models, making it one of the most powerful machine learning algorithms.

[0069] S4. Model building: Select model parameters for modeling, compare the weighted output results of the model under different parameters through cross-validation, then use the optimal parameters to build a regression model for the training set, and use the validation set to assess the model's capabilities. Depending on the size of the validation set, optimize the model parameters, and then regress the training set and validation set to finally complete the regression model building;

[0070] In step S4, when the model capability is assessed on the validation set, it is determined whether the root mean square error of the mechanical properties of the samples in the model prediction validation set is significantly higher than the root mean square error of the training set and the validation set, or whether the data-corrected coefficient of variation of the model prediction value and the validation set is significantly lower than the data-corrected coefficient of variation of the training set; if any one of the conditions is met, it means that the original data distribution has changed or the model selection is unreasonable, and the model needs to be rebuilt; if none of the conditions are met, the model parameters only need to be optimized according to the size of the validation set.

[0071] S5. Microstructure prediction: After completing the establishment of the regression model, the mass percentages of the independent variables C, Mn, Si, Cr, Mo, Ni, Cu and CE and t8 / 5 are obtained as input, and the volume fractions of the phase components are output.

[0072] S6. Mechanical properties prediction: A microstructure model is established based on the volume fraction of the phase components to predict the mechanical properties of the flux-cored welding wire.

[0073] Step S6 is specifically as follows: after obtaining the predicted results of the volume fractions of the phase components of the welding wire structure, a series of microstructure models that conform to the volume fractions of the phase components are first generated by computer based on the predicted results, and two-point statistical function analysis of the microstructure model is performed using software tools such as PSPSystem, and the microstructure model with the lowest residual is selected to establish a two-dimensional representative volume unit; then, the engineering stress-strain curve of the welding wire structure is calculated by the crystal plasticity method, and finally, the predicted values of the yield strength and tensile strength of the welding wire structure are obtained based on the calculated values of the engineering stress-strain curve.

[0074] Among them, PSPSystem (Pattern Spectrum System) is a software tool used to analyze the microstructural characteristics of materials. It quantifies the spatial distribution characteristics of the microstructure based on a two-point statistical function (also known as the radial distribution function or structure factor).

[0075] Example 1

[0076] In this embodiment, a method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel is proposed.

[0077] The prediction method comprises the following steps:

[0078] S1. Data division: Accumulate the quantitative analysis data of welding wire structure, and divide the accumulated quantitative analysis data of welding wire structure into a training set and a validation set.

[0079] The accumulated quantitative analysis data of the welding wire structure includes weld metal with a yield strength of 350-1020 MPa.

[0080] In step S1, the accumulated quantitative analysis data of the welding wire structure are the component analysis results, process parameters, and volume fractions of phase components corresponding to the welding wire structure under different welding processes.

[0081] The composition analysis results include the mass percentages of C, Mn, Si, Cr, Mo, Ni, Cu and CE.

[0082] The process parameters include t8 / 5.

[0083] The volume fractions of the phase components include the volume fraction of acicular ferrite AF, the volume fraction of proeutectoid ferrite PF, the volume fraction of bainite B and the volume fraction of martensite M.

[0084] S2. Data standardization: After data partitioning, the training set and validation set are subjected to Z-score standardization respectively;

[0085] In step S2, the calculation steps for Z-score standardization are: new data = (original data - mean) / standard deviation.

[0086] S3. Model selection: The GBDT Gradient Boosting Decision Tree method is selected to establish an organizational prediction model.

[0087] S4. Model establishment: Through feature screening, model parameters such as w(C), w(Mn), w(Si), w(Cr), w(Mo), w(Ni), w(V+Ti), w(Cu), and t8 / 5 are selected for modeling. The weighted output results of the model under different parameters are compared through cross-validation, and then the best parameters are used to establish a regression model for the training set, and the model capability is evaluated using the validation set. The results are as follows: Figure 3 As shown, from Figure 3 It can be seen that the predicted values of each phase component (acicular ferrite AF, proeutectoid ferrite PF, bainite B and martensite M) are consistent with the actual values, indicating that the model performs well on the validation set.

[0088] The model parameters are optimized according to the size of the validation set, and then regression is performed on the training set and validation set to finally complete the establishment of the regression model. The results are shown in Table 2.

[0089] Table 2 GBDT model parameters

[0090]

[0091] As can be seen from Table 2, the mean corrected coefficient of determination of the training set of each phase component (acicular ferrite AF, proeutectoid ferrite PF, bainite B and martensite M) is greater than 0.9, indicating that the model performs well on the training set; the mean square error of the validation set of each phase component (acicular ferrite AF, proeutectoid ferrite PF, bainite B and martensite M) is less than 37.5. As can be seen from Table 2, the mean corrected coefficient of determination of the validation set of each phase component (acicular ferrite AF, proeutectoid ferrite PF, bainite B and martensite M) is greater than 0.87, indicating that the model performs well on the validation set.

[0092] S5. Microstructure prediction: After completing the establishment of the regression model, the mass percentages of the independent variables C, Mn, Si, Cr, Mo, Ni, Cu and CE, namely w(C), w(Mn), w(Si), w(Cr), w(Mo), w(Ni), w(Cu) and w(CE) and t8 / 5 are obtained as input, and the volume fractions of the phase components are output, namely the volume fractions of acicular ferrite AF, proeutectoid ferrite PF, bainite B and martensite M.

[0093] S6. Mechanical properties prediction: A microstructure model is established based on the volume fraction of the phase components to predict the mechanical properties of the flux-cored welding wire.

[0094] Step S6 is specifically as follows: after obtaining the predicted results of the volume fractions of the phase components of the welding wire structure, a series of microstructure models that conform to the volume fractions of the phase components are first generated by computer based on the predicted results, and the two-point statistical function analysis of the microstructure model is performed using PSPSystem, and the microstructure model with the lowest residual is selected to establish a two-dimensional representative volume unit; then, the engineering stress-strain curve of the welding wire structure is calculated by the crystal plasticity method, and finally, the predicted values of the yield strength and tensile strength of the welding wire structure are obtained based on the calculated values of the engineering stress-strain curve.

[0095] In this embodiment, the structure of the weld metal is first predicted. The composition and process parameters of the weld metal are shown in Table 4. As shown in Table 4, w(C)=0.033%, w(Mn)=1.10%, w(Si)=0.22%, w(Cr)=0.25%, w(Mo)=0.4%, w(Ni)=0.7%, w(V)=0.04%, w(Cu)=0.20%, t8 / 5=9s, and w(CE)=0.414% is calculated. In this embodiment, when welding the weld metal, the welding current I is 210A, the welding voltage U is 30V, the welding speed v is 35cm / min, the corresponding heat input is 10kJ / cm, and the preheating temperature is 50°C.

[0096] The above independent variables were input into the regression model, and the output volume fractions of acicular ferrite AF, proeutectoid ferrite PF, bainite B and martensite M were shown in Table 5.

[0097] As shown in Table 5, the volume fraction of the output acicular ferrite AF was 77%, the volume fraction of the proeutectoid ferrite PF was 19%, and the volume fraction of bainite B was 4%.

[0098] Then the mechanical properties of the weld metal are predicted. On the above basis, after obtaining the predicted results of the volume fraction of the phase components of the welding wire structure, a series of microstructure models that conform to the volume fraction of the phase components are generated by computer according to the predicted results. The two-point statistical function analysis of the microstructure model is performed using PSPSystem, and the microstructure model with the lowest residual is selected to establish a two-dimensional representative volume unit; the engineering stress-strain curve of the material is then calculated using the crystal plasticity method. The results are shown in Figure 4 Finally, according to the calculated values of the engineering stress-strain curve, the predicted values of the yield strength and tensile strength of the welding wire structure were read out. The results are shown in Table 5.

[0099] Among them, proeutectoid ferrite, acicular ferrite, martensite, and bainite phases all have BCC structures, and the main material parameters are determined according to Table 3. Since bainite is a medium-temperature transformation product between ferrite and martensite, the parameters are taken as average values for estimation.

[0100] Table 3 Mechanical properties analysis material parameters of phase components

[0101]

[0102] Table 4 Mass percentage of weld metal components and process parameters

[0103]

[0104] Table 5 Volume fraction and mechanical properties of weld metal microstructure components

[0105]

[0106] from Figure 4 It can be seen that the model prediction value curve of stress in this embodiment has a high degree of coincidence with the actual test curve, indicating that the model prediction value curve of stress is well consistent with the actual test curve.

[0107] It can be seen from Table 5 that the prediction error of the predicted values of the yield strength and tensile strength of this embodiment is less than 1.53%, indicating that the model predicted values of the yield strength and tensile strength of this embodiment are consistent with the actual test values.

[0108] Example 2

[0109] In this embodiment, different from the embodiment 1, in step S5,

[0110] First, the structure of the weld metal is predicted. The composition of the weld metal and the process parameters are shown in Table 4. As shown in Table 4, w(C)=0.092%, w(Mn)=1.41%, w(Si)=0.27%, w(Cr)=0.26%, w(Mo)=0.3%, w(Ni)=1.44%, w(V)=0.02%, w(Cu)=0.25%, t8 / 5=12s, and w(CE)=0.556% is calculated.

[0111] In this embodiment, the preheating temperature of the weld metal during welding is 100°C.

[0112] As shown in Table 5, the volume fraction of the output acicular ferrite AF was 72%, the volume fraction of the proeutectoid ferrite PF was 3%, and the volume fraction of the bainite B was 25%.

[0113] Then the mechanical properties of the weld metal are predicted. Figure 5 Finally, according to the calculated values of the engineering stress-strain curve, the predicted values of yield strength and tensile strength were read out. The results are shown in Table 5.

[0114] from Figure 5It can be seen that the model prediction value curve of stress in this embodiment has a high degree of coincidence with the actual test value curve, indicating that the model prediction value curve of stress is well consistent with the actual test curve.

[0115] It can be seen from Table 5 that the prediction error of the predicted values of the yield strength and tensile strength of this embodiment is less than 3.79%, indicating that the prediction error of the yield strength and tensile strength of this embodiment is small and the prediction accuracy is high.

[0116] In summary, the prediction method is suitable for predicting the volume fraction of phase components in the microstructure of flux-cored welding wire with a yield strength of 460-590 MPa, the yield strength and tensile strength of the flux-cored welding wire.

[0117] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the scope of protection of the present invention should be based on the scope defined by the claims.

Claims

1. A method for predicting the microstructure characteristics and mechanical properties of flux-cored welding wire for hull steel, characterized in that: The prediction method is applicable to predicting the volume fraction of phase components in the microstructure of a flux-cored welding wire and the mechanical properties of the flux-cored welding wire, wherein the phase components include acicular ferrite AF, proeutectoid ferrite PF, bainite B, and martensite M. The prediction method comprises the following steps: S1. Data division: Accumulate the quantitative analysis data of the welding wire structure, and divide the accumulated quantitative analysis data of the welding wire structure into a training set and a validation set. The training set is used to directly build the regression model, and the validation set is used to evaluate the model during the construction process; S2. Data standardization: After data partitioning, the training set and validation set are subjected to Z-score standardization respectively; S3, model selection: select GBDT algorithm to build the model; S4. Model building: Select model parameters for modeling, compare the weighted output results of the model under different parameters through cross-validation, then use the optimal parameters to build a regression model for the training set, and use the validation set to assess the model's capabilities. Depending on the size of the validation set, optimize the model parameters, and then regress the training set and validation set to finally complete the regression model building; S5. Microstructure prediction: After the regression model is established, the mass percentages of the independent variables C, Mn, Si, Cr, Mo, Ni, Cu, and CE and t8 / 5 are obtained as inputs, and the volume fractions of the phase components are output; S6. Mechanical properties prediction: Establish a microstructure model based on the volume fraction of phase components to predict the mechanical properties of flux-cored welding wire; Step S6 specifically comprises: after obtaining the predicted results of the volume fractions of the phase components of the welding wire structure, firstly, generating a series of microstructure models that conform to the volume fractions of the phase components by computer based on the predicted results, and using software tools to perform two-point statistical function analysis on the microstructure models, selecting the microstructure model with the lowest residual to establish a two-dimensional representative volume unit; then, calculating the engineering stress-strain curve of the welding wire structure by a crystal plasticity method, and finally, obtaining the predicted values of the yield strength and tensile strength of the welding wire structure based on the calculated values of the engineering stress-strain curve; In step S1, the accumulated quantitative analysis data of the welding wire structure includes weld metal with a yield strength of 350 to 1020 MPa; In step S1, the chemical composition of the weld metal includes, by mass percentage, w(C)≤0.1%, w(Mn): 0.60-1.5%, w(Si)≤0.4%, w(Cr)≤0.4%, w(Mo)≤0.5%, w(Ni): 0.6-1.5%, w(V)≤0.05%, w(Cu)≤0.5%, w(S)≤0.0025%, w(P)≤0.0025%, and the remainder is Fe and unavoidable impurities; In step S1, the accumulated quantitative analysis data of the welding wire structure are the component analysis results, process parameters, and volume fractions of phase components corresponding to the welding wire structure under different welding processes; The composition analysis results include at least the mass percentages of C, Mn, Si, Cr, Mo, Ni, Cu and CE; The process parameters include at least t8 / 5; In step S1, the accumulated welding wire structure quantitative analysis data is preprocessed, and the data is divided using a leave-one-out method or a k-fold cross-validation method, where k is greater than or equal to 50; In step S2, the calculation steps of Z-score standardization are: new data = (original data - mean) / standard deviation.

Images