Nodular cast iron composition and process design method based on active learning strategy

Abstract

The invention discloses a nodular cast iron component and process design method based on an active learning strategy, and belongs to the technical field of material process design. In order to improve the prediction precision and design efficiency of material process design, the method comprises the following steps: collecting experimental data including nodular cast iron components, process, microstructure characteristics and performance in literatures; constructing a nodular cast iron component-process-microstructure characteristic model and a nodular cast iron component-process-microstructure characteristic-performance model, setting different weights for different target performances based on an optimization algorithm, and obtaining optimized nodular cast iron components and process parameters with different performance requirements by adopting a weight-based target function; and through an active learning strategy, data with relatively large uncertainty is preferentially selected for experiment and is supplemented to a data set, then model updating is performed, model prediction precision is optimized, and finally nodular cast iron components and process parameters meeting expected performance requirements are designed. The experiment frequency and cost are reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of material process design, and in particular relates to a ductile iron composition and process design method based on an active learning strategy. Background Art

[0002] In recent years, driven by domestic energy demand and industrial needs, large-scale cast iron construction has received special attention. Ductile iron, as an important casting material, has good strength, toughness and fatigue resistance. It is widely used in automobiles, construction, pipelines and other fields, and is often used as a key component of a variety of cast iron castings. Therefore, it is very popular. Many equipment usually operate in very harsh environments. Therefore, designing ductile iron that meets various performance requirements has become a key issue for related industries.

[0003] However, the traditional ductile iron design method relies almost entirely on experience and repeated experimental adjustments. This method is not only time-consuming and labor-intensive, but also difficult to quickly optimize material properties because it is impossible to accurately grasp the interaction between various parameters. Therefore, the existing ductile iron design mainly faces the following problems:

[0004] 1. Low data utilization: The design process based on data modeling usually relies on a large amount of experimental data, but these data are often unevenly distributed. Under the existing technical methods, the combination of certain key parameters is not fully covered, and the design results are not ideal.

[0005] 2. Incomplete design content: Traditional design methods hardly involve the comprehensive composition-process-structure-performance relationship, but only involve several parts of it, which makes it impossible to effectively predict the structure and performance of cast iron under different compositions and processes, resulting in a slow process and insufficient design content.

[0006] 3. Unable to cope with multi-objective requirements: Existing design methods are difficult to take into account multiple performance objectives at the same time. For example, the balance requirements of tensile strength, toughness and hardness in different application scenarios often require a large number of experiments for verification, which increases the complexity and cost of the design.

[0007] 4. Lack of feedback optimization mechanism: Existing material design methods cannot be dynamically optimized through feedback from experiments that have already been conducted, resulting in deviations between model predictions and actual results, and are unable to effectively reduce design errors. Summary of the invention

[0008] The problem to be solved by the present invention is to improve the prediction accuracy and design efficiency of material process design, and propose a ductile iron composition and process design method based on an active learning strategy.

[0009] To achieve the above object, the present invention is implemented through the following technical solutions:

[0010] A method for designing ductile iron composition and process based on active learning strategy comprises the following steps:

[0011] S1. Collect experimental data including ductile iron composition, process, microstructure characteristics and properties in the literature, add the associated data with ductile iron composition, process, microstructure characteristics and properties in the experimental data to data set 1, and add the associated data with only ductile iron composition, process and properties in the experimental data to data set 2;

[0012] S2. Visualize the distribution of ductile iron composition or process data in data set 1, and then perform an orthogonal test based on the blank area of ​​the visualized ductile iron composition or process data to obtain orthogonal test data and add it to data set 1;

[0013] S3. Preprocess the data in data set 1 and data set 2 to obtain preprocessed data set 1 and preprocessed data set 2;

[0014] S4. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure feature model is constructed, the ductile iron composition and process are set as input, and the microstructure feature is used as output, and the ductile iron composition-process-microstructure feature model is trained using the preprocessed data set 1, and the preprocessed data set 2 is predicted using the trained ductile iron composition-process-microstructure feature model to obtain the microstructure feature corresponding to each group of data in the preprocessed data set 2, and the obtained microstructure feature data is normalized and added to the preprocessed data set 2;

[0015] S5. The preprocessed data set 1 and the preprocessed data set 2 obtained in step S4 are combined into data set 3;

[0016] S6. Based on the small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristics-performance model is constructed, the ductile iron composition, process and microstructure characteristics are set as input, and the performance is used as output. The ductile iron composition-process-microstructure characteristics-performance model is trained using data set 3 to obtain a trained ductile iron composition-process-microstructure characteristics-performance model;

[0017] S7. Based on the optimization algorithm, different weights are set for different target performances. Each target performance is calculated using the ductile iron composition-process-microstructure characteristics-performance model trained in step S6. The weight-based objective function is used to continuously select the components closest to the target performance to obtain optimized ductile iron compositions and process parameters with different performance requirements;

[0018] S8. For the optimized ductile iron composition and process parameters with different performance requirements obtained in step S7, an active learning strategy is used to prioritize data with larger uncertainty for experiments and add them to the data set, and then return to step S4 to update the model, optimize the model prediction accuracy, and finally design the ductile iron composition and process parameters that meet the expected performance requirements.

[0019] Furthermore, the data of the composition of ductile iron in step S1 mainly include carbon, silicon, manganese, rare earth, sulfur, phosphorus, copper, nickel, magnesium and chromium, the data of the process mainly include cooling rate, type of inoculant, content of inoculant, type of spheroidizer, content of spheroidizer, incubation temperature, incubation time, the data of microstructure characteristics mainly include graphite nodule diameter, number of graphite nodules, graphite spheroidization rate, proportion of pearlite in matrix, proportion of ferrite in matrix, spheroidization grade, and the data of performance mainly include tensile strength, yield strength, ductility and impact toughness.

[0020] Furthermore, in step S2, the first two bits with the most microstructure feature data in data set 1 are selected as the output features of the model, or the microstructure features in data set 1 are customized as the output features of the model, and the distribution of ductile iron composition or process data associated with the selected or customized microstructure feature data is visualized to extract blank areas.

[0021] Furthermore, the preprocessing in step S3 includes data cleaning, normalization processing and encoding processing. The formula for normalization processing is:

[0022]

[0023] Among them, X is the original data, X min and X max are the minimum and maximum values ​​in the data set, respectively, and X' is the normalized data;

[0024] The coding process is to code the type of inoculant and the type of spheroidizing agent.

[0025] Furthermore, the specific implementation method of step S4 includes the following steps:

[0026] S4.1. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristic model is constructed, and the ductile iron composition and process are set as input, and the microstructure characteristics are set as output. The small sample machine learning algorithm includes an extreme random tree algorithm, a CatBoost algorithm, a random forest algorithm, and an adaptive boost regression algorithm;

[0027] S4.2. Set the performance of the model to be evaluated by methods such as k-fold cross validation, k = 5-10, and select the best performing ductile iron composition-process-microstructure characteristic model. When predicting multiple microstructure characteristics, adopt the strategy of establishing models separately;

[0028] S4.3. Set evaluation criteria including mean absolute error MAE, root mean square error RMSE, and determination coefficient R 2 And the average relative error MAPE, calculated as:

[0029]

[0030]

[0031]

[0032]

[0033] in, Represents the predicted value of the machine learning model, y i represents the true value;

[0034] Then the algorithm with the best evaluation criterion is selected as the final trained ductile iron composition-processing-microstructure characteristic model.

[0035] Furthermore, in the extreme random tree algorithm in step S4.1, the number of trees is 100, the maximum depth is 10, and the minimum number of samples required for splitting each node is 2; in the CatBoost algorithm, the number of iterations training rounds is 300, the learning rate is 0.02, and the L2 regularization coefficient is 3; in the random forest algorithm, the number of trees in the forest is 100, the minimum number of split samples is 2, and the minimum number of samples for leaf nodes is 2; in the adaptive boosting regression algorithm, the number of basic models is 80, the learning rate is 0.02, and the loss function type is linear.

[0036] Furthermore, the specific implementation method of step S6 includes the following steps:

[0037] S6.1. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristics-performance model is constructed, and the ductile iron composition, process, and microstructure characteristics are set as inputs, and the performance is used as output. The small sample machine learning algorithm includes an extreme random tree algorithm, a CatBoost algorithm, a random forest algorithm, and an adaptive boost regression algorithm;

[0038] S6.2. Set up a k-fold cross validation method to evaluate the performance of the model, k = 5-10, and select the best performing ductile iron composition-processing-microstructure characteristics-performance model. When predicting multiple performance characteristics, adopt a strategy of establishing models separately;

[0039] S6.3. Set evaluation criteria including mean absolute error (MAE), root mean square error (RMSE), and determination coefficient (R) 2 And the mean relative error MAPE.

[0040] Furthermore, in step S7, the particle swarm optimization algorithm is used to perform intelligent optimization of the components and processes. The objective function f of the particle swarm optimization algorithm is defined as:

[0041]

[0042] Among them, K i is the weight of the i-th target, T i is the target value, T' i is the model's predicted value.

[0043] Furthermore, the calculation method of the uncertainty in step S8 is: when other conditions remain unchanged, the hyperparameters of the composition-process-microstructure characteristics-performance model are changed, and the performance prediction values ​​of the composition and process optimized by the particle swarm optimization algorithm in the previous step are calculated respectively. The uncertainty U is the variance of the prediction values ​​of different models, and the expression is:

[0044]

[0045] in, is the average prediction of all models, is the predicted value of the mth model, and M is the total number of models.

[0046] Beneficial effects of the present invention:

[0047] The composition and process design method of ductile iron based on active learning strategy described in the present invention realizes dynamic optimization of model performance through experimental feedback by introducing active learning strategy; covers comprehensive composition-process-structure-performance relationship through step-by-step method; solves the problem of uneven data distribution through supplementary sampling, and effectively improves the prediction accuracy and design efficiency of the model. At the same time, the method can intelligently optimize the composition ratio and process parameters of ductile iron according to a variety of different design goals (such as tensile strength, impact toughness, etc.), effectively serving the actual application needs.

[0048] The ductile iron composition and process design method based on active learning strategy described in the present invention improves the ductile iron design efficiency, reduces the number of experiments and costs, solves many drawbacks of existing data modeling design methods, and has good promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flow chart of a ductile iron composition and process design method based on an active learning strategy according to the present invention;

[0050] Figure 2 It is a visualization diagram of the data distribution of C, Si and Mn in Example 2 of the present invention;

[0051] Figure 3 This is a visualization diagram of the data distribution of Cu, Ni and Re in Example 2 of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present invention more clear, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the specific embodiments described are only part of the embodiments of the present invention, rather than all of the specific embodiments. The components of the specific embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations, and the present invention can also have other embodiments.

[0053] Therefore, the following detailed description of the specific embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents the selected specific embodiments of the present invention. Based on the specific embodiments of the present invention, all other specific embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0054] In order to further understand the content, features and effects of the present invention, the following specific implementation methods are given as examples, and the attached Figure 1 -Attached Figure 3 The detailed instructions are as follows:

[0055] Embodiment 1:

[0056] A method for designing ductile iron composition and process based on active learning strategy comprises the following steps:

[0057] S1. Collect experimental data including ductile iron composition, process, microstructure characteristics and properties in the literature, add the associated data with ductile iron composition, process, microstructure characteristics and properties in the experimental data to data set 1, and add the associated data with only ductile iron composition, process and properties in the experimental data to data set 2;

[0058] Further, the data of the ductile iron composition in step S1 mainly include carbon, silicon, manganese, rare earth, sulfur, phosphorus, copper, nickel, magnesium and chromium, the data of the process mainly include cooling rate, inoculant type, inoculant content, spheroidizing agent type, spheroidizing agent content, incubation temperature, incubation time, the data of microstructure characteristics mainly include graphite ball diameter, graphite ball number, graphite spheroidization rate, pearlite ratio in matrix, ferrite ratio in matrix, spheroidization grade, and the data of performance mainly include tensile strength, yield strength, ductility and impact toughness;

[0059] S2. Visualize the distribution of ductile iron composition or process data in data set 1, and then perform an orthogonal test based on the blank area of ​​the visualized ductile iron composition or process data to obtain orthogonal test data and add it to data set 1;

[0060] Further, in step S2, the first two bits with the most microstructure feature data in data set 1 are selected as the output features of the model or the microstructure features in the customized data set 1 are used as the output features of the model, and the ductile iron composition or process associated with the selected or customized microstructure feature data is visualized to extract the blank area;

[0061] S3. Preprocess the data in data set 1 and data set 2 to obtain preprocessed data set 1 and preprocessed data set 2;

[0062] Furthermore, the preprocessing in step S3 includes data cleaning, normalization processing and encoding processing. The formula for normalization processing is:

[0063]

[0064] Among them, X is the original data, X min and X max are the minimum and maximum values ​​in the data set, respectively, and X' is the normalized data;

[0065] The coding process is to code the types of inoculants and spheroidizers;

[0066] S4. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure feature model is constructed, the ductile iron composition and process are set as input, and the microstructure feature is used as output, and the ductile iron composition-process-microstructure feature model is trained using the preprocessed data set 1, and the preprocessed data set 2 is predicted using the trained ductile iron composition-process-microstructure feature model to obtain the microstructure feature corresponding to each group of data in the preprocessed data set 2, and the obtained microstructure feature data is normalized and added to the preprocessed data set 2;

[0067] Furthermore, the specific implementation method of step S4 includes the following steps:

[0068] S4.1. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristic model is constructed, and the ductile iron composition and process are set as input, and the microstructure characteristics are set as output. The small sample machine learning algorithm includes an extreme random tree algorithm, a CatBoost algorithm, a random forest algorithm, and an adaptive boost regression algorithm;

[0069] Furthermore, in the extreme random tree algorithm in step S4.1, the number of trees is 100, the maximum depth is 10, and the minimum number of samples required for each node split is 2; in the CatBoost algorithm, the number of iterations training rounds is 300, the learning rate is 0.02, and the L2 regularization coefficient is 3; in the random forest algorithm, the number of trees in the forest is 100, the minimum number of split samples is 2, and the minimum number of samples for leaf nodes is 2; in the adaptive boosting regression algorithm, the number of basic models is 80, the learning rate is 0.02, and the loss function type is linear;

[0070] Furthermore, the random forest is composed of multiple decision trees, and the final output is the mean of the predictions of each tree. The output of the random forest can be expressed as:

[0071]

[0072] Among them, Y i (x) represents the prediction result of the i-th tree, and N is the number of decision trees.

[0073] AdaBoost trains the model by updating the learner weights, and its model output is:

[0074]

[0075] Among them, α t is the weight of the t-th base learner, h t (x) is the prediction result of the tth weak learner.

[0076] CatBoost uses gradient boosted decision trees for regression to minimize the loss function, which is expressed as:

[0077]

[0078] S4.2. Set the performance of the model to be evaluated by methods such as k-fold cross validation, k = 5-10, and select the best performing ductile iron composition-process-microstructure characteristic model. When predicting multiple microstructure characteristics, adopt the strategy of establishing models separately;

[0079] S4.3. Set evaluation criteria including mean absolute error MAE, root mean square error RMSE, and determination coefficient R 2 And the average relative error MAPE, calculated as:

[0080]

[0081] in, Represents the predicted value of the machine learning model, y i represents the true value;

[0082] Then the algorithm with the best evaluation criteria is selected as the final trained ductile iron composition-process-microstructure feature model

[0083] S5. The preprocessed data set 1 and the preprocessed data set 2 obtained in step S4 are combined into data set 3;

[0084] S6. Based on the small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristics-performance model is constructed, the ductile iron composition, process and microstructure characteristics are set as input, and the performance is used as output. The ductile iron composition-process-microstructure characteristics-performance model is trained using data set 3 to obtain a trained ductile iron composition-process-microstructure characteristics-performance model;

[0085] Furthermore, the specific implementation method of step S6 includes the following steps:

[0086] S6.1. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristics-performance model is constructed, and the ductile iron composition, process, and microstructure characteristics are set as inputs, and the performance is used as output. The small sample machine learning algorithm includes an extreme random tree algorithm, a CatBoost algorithm, a random forest algorithm, and an adaptive boost regression algorithm;

[0087] Furthermore, in the extreme random tree algorithm, n_estimators (the number of trees) is 100, the maximum depth is 10, and the minimum number of samples required for each node split is 2; in the CatBoost algorithm, the number of iterations training rounds is 300, the learning_rate learning rate is 0.02, and the L2 regularization coefficient is 3; in the random forest algorithm, n_estimators: the number of trees in the forest is, 00, the minimum number of split samples is 2, and the minimum number of samples for leaf nodes is 2; in the adaptive boosting regression algorithm, n_estimators: the number of basic models is 80, the learning_rate learning rate is 0.02, and the loss function type is 'linear';

[0088] S6.2. Set up a k-fold cross validation method to evaluate the performance of the model, k = 5-10, and select the best performing ductile iron composition-processing-microstructure characteristics-performance model. When predicting multiple performance characteristics, adopt a strategy of establishing models separately;

[0089] S6.3. Set evaluation criteria including mean absolute error (MAE), root mean square error (RMSE), and determination coefficient (R) 2 and mean relative error MAPE;

[0090] S7. Based on the optimization algorithm, different weights are set for different target performances. Each target performance is calculated using the ductile iron composition-process-microstructure characteristics-performance model trained in step S6. The weight-based objective function is used to continuously select the components closest to the target performance to obtain optimized ductile iron compositions and process parameters with different performance requirements;

[0091] Furthermore, in step S7, the particle swarm optimization algorithm is used to perform intelligent optimization of the components and processes. The objective function f of the particle swarm optimization algorithm is defined as:

[0092]

[0093] Among them, K i is the weight of the i-th target, T i is the target value, T' i is the model prediction value;

[0094] S8. For the optimized ductile iron composition and process parameters with different performance requirements obtained in step S7, an active learning strategy is used to prioritize data with larger uncertainty for experiments and add them to the data set, and then return to step S4 to update the model, optimize the model prediction accuracy, and finally design the ductile iron composition and process parameters that meet the expected performance requirements.

[0095] Furthermore, the calculation method of the uncertainty in step S8 is: when other conditions remain unchanged, the hyperparameters of the composition-process-microstructure characteristics-performance model are changed, and the performance prediction values ​​of the composition and process optimized by the particle swarm optimization algorithm in the previous step are calculated respectively. The uncertainty U is the variance of the prediction values ​​of different models, and the expression is:

[0096]

[0097] in, is the average prediction of all models, is the predicted value of the mth model, and M is the total number of models.

[0098] Embodiment 2:

[0099] Based on Example 1, the target ductile iron performance is set to a tensile strength greater than 550 MPa and a V-notch impact energy greater than 12 J at room temperature, and a detailed example is given as follows:

[0100] This embodiment solves the problems of low data utilization, incomplete design content, inability to cope with multi-objective design requirements and lack of feedback optimization mechanism in existing data modeling design methods.

[0101] S1. Collect experimental data including ductile iron composition, process, microstructure characteristics and properties in the literature, add the associated data with ductile iron composition, process, microstructure characteristics and properties in the experimental data to data set 1, and add the associated data with only ductile iron composition, process and properties in the experimental data to data set 2;

[0102] Furthermore, the target ductile iron properties are set to tensile strength greater than 550MPa and V-notch impact energy greater than 12J at room temperature. First, 202 sets of ductile iron data containing organizational information are collected from databases such as CNKI using "ductile iron" as the keyword as data set 1 and 67 sets of data without organizational information are collected as data set 2. The data cover the chemical composition, smelting process, organizational appearance and various mechanical properties of different ductile irons under the same environmental conditions. The detailed data types are shown in Table 1:

[0103] Table 1 Ductile iron data set to be collected

[0104]

[0105] S2. Visualize the distribution of ductile iron composition or process data in data set 1, and then perform an orthogonal test based on the blank area of ​​the visualized ductile iron composition or process data to obtain orthogonal test data and add it to data set 1;

[0106] Further, orthogonal experimental design: Based on data set 1, analyze and find blank or sparse areas of certain chemical compositions or process parameters. Visualize the distribution of carbon-manganese-silicon composition data and copper-nickel-rare earth composition data in the data set. Figure 2 and Figure 3As shown. It can be found that the data distribution is uneven for carbon elements in the range of 2.2-2.4wt%, silicon elements in the range of 3.0-4.3wt%, manganese elements in the range of 0.5-0.8wt%, and rare earth elements in the range of 0.02-0.05wt%. In order to make up for these blank areas and improve the data quality, this embodiment uses an orthogonal experiment for supplementary sampling. The gradients of carbon, silicon, manganese, and rare earth components are set to 0.1, 0.2, 0.1, and 0.01, respectively, and an orthogonal experiment is designed. The smelting process adopts the impact spheroidization method, the spheroidization temperature is 1450°C, the spheroidization treatment time is 60s, and rare earth magnesium alloy spheroidizer and FeSi75 inoculant are used. The data obtained from the experiment are supplemented to the data set.

[0107] S3. Preprocess the data in data set 1 and data set 2 to obtain preprocessed data set 1 and preprocessed data set 2;

[0108] Furthermore, necessary preprocessing is performed on the collected experimental data to ensure data quality and consistency. Normalization: variables of different scales (such as graphite spheroidization rate and tensile strength) are converted into a unified [0,1] range.

[0109] For some classification features (such as inoculant type and spheroidizer type), coding processing is performed. After preliminary processing to delete invalid, abnormal and duplicate data, the ductile iron data set has a total of 218 data sets, of which 37 data sets do not contain organizational information.

[0110] S4. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure feature model is constructed, the ductile iron composition and process are set as input, and the microstructure feature is used as output, and the ductile iron composition-process-microstructure feature model is trained using the preprocessed data set 1, and the preprocessed data set 2 is predicted using the trained ductile iron composition-process-microstructure feature model to obtain the microstructure feature corresponding to each group of data in the preprocessed data set 2, and the obtained microstructure feature data is normalized and added to the preprocessed data set 2;

[0111] Furthermore, the performance of different algorithms is shown in Table 2 and Table 3:

[0112] Table 2 Composition-process-spheroidization grade model indicators

[0113] Model MAE RMSE R2 percentage% ExtraTrees 0.3332 0.5071 0.7666 14.77 CatBoost 0.3448 0.5172 0.7498 15.52 RandomForest 0.374 0.5471 0.6839 16.81 AdaBoostRegressor 0.4406 0.5453 0.6739 19.64

[0114] Table 3 Composition-processing-pearlite content model index

[0115] Model MAE RMSE <![CDATA[R 2 ]]> percentage% ExtraTrees 4.6619 6.2455 0.8217 21.63 CatBoost 5.4326 7.387 0.7024 26.08 AdaBoostRegressor 6.1398 7.8395 0.6631 38.82 RandomForest 5.849 7.8791 0.6341 29.73

[0116] By comparing different algorithms, the Extra Trees algorithm with the best performance in various indicators was selected as the subsequent algorithm. For the 37 groups of data in the data set that do not contain organizational information, the trained composition-process-organization model was used to predict the missing organizational information. The predicted organizational data was added to the data set and used as the input data for the next step together with the original 218 groups of data.

[0117] S5. The preprocessed data set 1 and the preprocessed data set 2 obtained in step S4 are combined into data set 3;

[0118] S6. Based on the small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristics-performance model is constructed, the ductile iron composition, process and microstructure characteristics are set as input, and the performance is used as output. The ductile iron composition-process-microstructure characteristics-performance model is trained using data set 3 to obtain a trained ductile iron composition-process-microstructure characteristics-performance model;

[0119] Furthermore, the optimal algorithm is the Extra Trees algorithm. The performance of the two performance models is shown in Table 4:

[0120] Table 4 Optimal composition-process-structure-performance model indicators

[0121] Target MAE RMSE <![CDATA[R 2 ]]> percentage% Impact energy / J 1.6027 1.9027 0.8400 15.99 Tensile strength / MPa 22.0486 33.1265 0.9356 5.90

[0122] S7. Based on the optimization algorithm, different weights are set for different target performances. Each target performance is calculated using the ductile iron composition-process-microstructure characteristics-performance model trained in step S6. The weight-based objective function is used to continuously select the components closest to the target performance to obtain optimized ductile iron compositions and process parameters with different performance requirements;

[0123] Furthermore, the optimal machine learning model and particle swarm optimization algorithm were used to optimize the composition-process that meets the performance target. The target was set to 550MPa tensile strength and 12J V-notch impact energy at room temperature. The performance weights were 0.4 and 0.6 respectively. The objective function in the optimization algorithm was set to the weighted average of different targets, that is:

[0124] f=0.4*|T′ 1 -550|+0.6*|T′ 2 -12|

[0125] Among them, T' 1 and T' 2 They are the tensile strength predicted by the model and the V-notch impact energy at room temperature. By continuously iteratively adjusting the chemical composition and process parameters, the eight components and process combinations with the smallest objective function were found.

[0126] S8. For the optimized ductile iron composition and process parameters with different performance requirements obtained in step S7, an active learning strategy is used to prioritize data with larger uncertainty for experiments and add them to the data set, and then return to step S4 to update the model, optimize the model prediction accuracy, and finally design the ductile iron composition and process parameters that meet the expected performance requirements.

[0127] Furthermore, for the 8 groups of components and processes optimized by the particle swarm optimization algorithm in the previous step, the uncertainty of their predicted values ​​is calculated, and data with larger uncertainty are selected for experiments. The calculation method of uncertainty is: when other conditions remain unchanged, change the hyperparameters of the component-process-organization-performance model, that is, keep the maximum depth at 10, and the minimum number of samples required for each node splitting at 2. In the extreme random tree algorithm, n_estimators are 100, 140, 180, 220, and 260, respectively, and the performance prediction values ​​of the 8 groups of components and processes are calculated respectively, and the uncertainty U (the variance of the predicted values ​​of different models) is calculated respectively:

[0128]

[0129] in, is the average prediction of all models, is the predicted value of the mth model.

[0130] The three groups of components and processes with the largest uncertainty were screened out respectively, and the results are shown in Table 5:

[0131] Table 5 Screened components with the largest uncertainty in predicted values ​​- process

[0132] C Si Mn … Mg RE Technology Uncertainty type 3.92 2.48 0.17 … 0.056 0.043 … 9.4 Impact toughness 3.84 2.46 0.17 … 0.058 0.039 … 8.8 Impact toughness 3.78 2.21 0.25 … 0.062 0.021 … 7.1 Impact toughness 2.23 3.76 0.33 … 0.007 0.009 … 836.44 tensile strength 2.29 3.09 0.64 … 0.033 0.004 … 817.5 tensile strength 3.68 3.21 0.62 … 0.013 0.043 … 786.60 tensile strength

[0133] The screened components and processes are experimentally verified, and the obtained components, processes, organization and performance data are recorded. The components, processes, organization and performance data obtained after the experiment are added to the data set, and the two-step model is rebuilt, corrected and updated to improve the prediction accuracy and form a closed-loop optimization process of data-model-experiment.

[0134] Furthermore, the final model with high accuracy optimized based on active learning strategy was used to optimize the composition-process to meet the performance target again. The target was set to 550MPa tensile strength and 12J V-notch impact energy at room temperature, with performance weights of 0.4 and 0.6 respectively. Finally, the ductile iron composition and smelting process parameters that met the target performance were designed and experimentally verified. The experiment showed that the designed as-cast ductile iron had a tensile strength of 550MPa and a V-notch impact energy of 12J at room temperature.

[0135] Embodiment 3:

[0136] The difference between this embodiment and embodiment 2 is that in the two-step machine learning model building step, the 5-fold cross-validation method is not used, but the data set is divided into two parts, 85% of which is used as a training set and the other 15% is used as a test set. The algorithms used include support vector machine algorithm, decision tree algorithm, and logistic regression. In order to evaluate the performance of different models, the root mean square error (RMSE) and the coefficient of determination (R 2 ) and mean relative error (MAPE) were used as evaluation criteria.

[0137] Embodiment 4:

[0138] The difference between this embodiment and embodiment 2 is that the target ductile iron performance is a tensile strength of 500MPa and an elongation at break greater than 18%. In the data collection step, the collected data covers the chemical composition, smelting process, organizational morphology and various mechanical properties of different ductile irons under the same environmental conditions. Specifically, the composition (wt.%) includes C, Si, Mn, P, S, Cu, RE, Ni, Cr, Mg; the smelting process includes the type of spheroidizer, the type of inoculant, and the incubation time s; the organizational characteristics include pearlite content %, spheroidization grade, and graphite ball diameter μm; the mechanical properties include tensile strength MPa and elongation at break %. The final designed ductile iron was tested to have a tensile strength of 501MPa, a yield strength of 331MPa, and an elongation at break of 21%.

[0139] Embodiment 5:

[0140] The difference between this embodiment and embodiment 2 is that the target ductile iron properties are tensile strength of 460 MPa and V-notch impact value of 7.0 J at -40°C. In the data collection step, the mechanical properties include tensile strength MPa and V-notch impact value at -40°C. The final designed ductile iron was tested to have a tensile strength of 458 MPa and a V-notch impact value of 6.8 J at -40°C.

[0141] It should be noted that relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0142] Although the present application has been described above with reference to specific embodiments, various modifications may be made thereto and parts thereof may be replaced with equivalents without departing from the scope of the present application. In particular, as long as there is no structural conflict, the various features in the specific embodiments disclosed in the present application may be used in combination with each other in any manner, and the fact that these combinations are not exhaustively described in this specification is only for the sake of omitting space and saving resources. Therefore, the present application is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A method for designing ductile iron composition and process based on active learning strategy, characterized in that: The steps include: S1. Collect experimental data including ductile iron composition, process, microstructure characteristics and properties in the literature, add the associated data with ductile iron composition, process, microstructure characteristics and properties in the experimental data to data set 1, and add the associated data with only ductile iron composition, process and properties in the experimental data to data set 2; S2. Visualize the distribution of ductile iron composition or process data in data set 1, and then perform an orthogonal test based on the blank area of ​​the visualized ductile iron composition or process data to obtain orthogonal test data and add it to data set 1; S3. Preprocess the data in data set 1 and data set 2 to obtain preprocessed data set 1 and preprocessed data set 2; S4. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure feature model is constructed, the ductile iron composition and process are set as input, and the microstructure feature is used as output, and the ductile iron composition-process-microstructure feature model is trained using the preprocessed data set 1, and the preprocessed data set 2 is predicted using the trained ductile iron composition-process-microstructure feature model to obtain the microstructure feature corresponding to each group of data in the preprocessed data set 2, and the obtained microstructure feature data is normalized and added to the preprocessed data set 2; S5. The preprocessed data set 1 and the preprocessed data set 2 obtained in step S4 are combined into data set 3; S6. Based on the small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristics-performance model is constructed, the ductile iron composition, process and microstructure characteristics are set as input, and the performance is used as output. The ductile iron composition-process-microstructure characteristics-performance model is trained using data set 3 to obtain a trained ductile iron composition-process-microstructure characteristics-performance model; S7. Based on the optimization algorithm, different weights are set for different target performances. Each target performance is calculated using the ductile iron composition-process-microstructure characteristics-performance model trained in step S6. The weight-based objective function is used to continuously select the components closest to the target performance to obtain optimized ductile iron compositions and process parameters with different performance requirements; S8. For the optimized ductile iron composition and process parameters with different performance requirements obtained in step S7, an active learning strategy is used to prioritize data with larger uncertainty for experiments and add them to the data set, and then return to step S4 to update the model, optimize the model prediction accuracy, and finally design the ductile iron composition and process parameters that meet the expected performance requirements.

2. A ductile iron composition and process design method based on active learning strategy according to claim 1, characterized in that: The data of the ductile iron composition in step S1 mainly include carbon, silicon, manganese, rare earth, sulfur, phosphorus, copper, nickel, magnesium and chromium, the data of the process mainly include cooling rate, type of inoculant, content of inoculant, type of spheroidizing agent, content of spheroidizing agent, incubation temperature, incubation time, the data of microstructure characteristics mainly include graphite nodule diameter, number of graphite nodules, graphite spheroidization rate, pearlite ratio in matrix, ferrite ratio in matrix, spheroidization grade, and the data of performance mainly include tensile strength, yield strength, ductility and impact toughness.

3. A ductile iron composition and process design method based on active learning strategy according to claim 2, characterized in that: In step S2, the first two bits with the most microstructure feature data in data set 1 are selected as the output features of the model, or the microstructure features in the customized data set 1 are used as the output features of the model, and the distribution of ductile iron composition or process data associated with the selected or customized microstructure feature data is visualized to extract blank areas.

4. A method for designing ductile iron composition and process based on active learning strategy according to claim 3, characterized in that: The preprocessing in step S3 includes data cleaning, normalization and encoding. The formula for normalization is: Among them, X is the original data, X min and X max are the minimum and maximum values ​​in the data set, respectively, and X' is the normalized data; The coding process is to code the type of inoculant and the type of spheroidizing agent.

5. A method for designing ductile iron composition and process based on active learning strategy according to claim 4, characterized in that: The specific implementation method of step S4 includes the following steps: S4.

1. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristic model is constructed, and the ductile iron composition and process are set as input, and the microstructure characteristics are set as output. The small sample machine learning algorithm includes an extreme random tree algorithm, a CatBoost algorithm, a random forest algorithm, and an adaptive boost regression algorithm; S4.

2. Set the performance of the model to be evaluated by methods such as k-fold cross validation, k = 5-10, and select the best performing ductile iron composition-process-microstructure characteristic model. When predicting multiple microstructure characteristics, adopt the strategy of establishing models separately; S4.

3. Set evaluation criteria including mean absolute error (MAE), root mean square error (RMSE), and determination coefficient (R) 2 And the average relative error MAPE, calculated as: in, Represents the predicted value of the machine learning model, y i represents the true value; Then the algorithm with the best evaluation criterion is selected as the final trained ductile iron composition-processing-microstructure characteristic model.

6. A method for designing ductile iron composition and process based on active learning strategy according to claim 5, characterized in that: In the extreme random tree algorithm in step S4.1, the number of trees is 100, the maximum depth is 10, and the minimum number of samples required for each node split is 2; in the CatBoost algorithm, the number of iterations training rounds is 300, the learning rate is 0.02, and the L2 regularization coefficient is 3; in the random forest algorithm, the number of trees in the forest is 100, the minimum number of split samples is 2, and the minimum number of samples for leaf nodes is 2; In the adaptive boosting regression algorithm, the number of base models is 80, the learning rate is 0.02, and the loss function type is linear.

7. A method for designing ductile iron composition and process based on active learning strategy according to claim 6, characterized in that: The specific implementation method of step S6 includes the following steps: S6.

1. Based on a small sample machine learning algorithm, a ductile iron composition-process-microstructure characteristics-performance model is constructed, and the ductile iron composition, process, and microstructure characteristics are set as inputs, and the performance is used as output. The small sample machine learning algorithm includes an extreme random tree algorithm, a CatBoost algorithm, a random forest algorithm, and an adaptive boost regression algorithm; S6.

2. Set up a k-fold cross validation method to evaluate the performance of the model, k = 5-10, and select the best performing ductile iron composition-processing-microstructure characteristics-performance model. When predicting multiple performance characteristics, adopt a strategy of establishing models separately; S6.

3. Set evaluation criteria including mean absolute error (MAE), root mean square error (RMSE), and determination coefficient (R) 2 And the mean relative error MAPE.

8. A method for designing ductile iron composition and process based on active learning strategy according to claim 7, characterized in that: In step S7, the particle swarm optimization algorithm is used to perform intelligent optimization of components and processes. The objective function f of the particle swarm optimization algorithm is defined as: Among them, K i is the weight of the i-th target, T i is the target value, T' i is the model's predicted value.

9. A method for designing ductile iron composition and process based on active learning strategy according to claim 7, characterized in that: The calculation method of uncertainty in step S8 is: when other factors remain unchanged, change the hyperparameters of the composition-process-microstructure characteristics-performance model, and calculate the performance prediction values ​​of the composition and process optimized by the particle swarm optimization algorithm in the previous step respectively. The uncertainty U is the variance of the prediction values ​​of different models, and the expression is: in, is the average prediction of all models, is the predicted value of the mth model, and M is the total number of models.

Images