Prediction method for oxidation weight increment of gamma '-phase reinforced cobalt-based high-temperature alloy

Abstract

The invention belongs to the technical field of cobalt-based high-temperature alloys, and particularly discloses a gamma '-phase reinforced cobalt-based high-temperature alloy oxidation weight increment prediction method which comprises the following steps: collecting a characteristic data set for predicting the average oxidation rate of a cobalt-based high-temperature alloy; preprocessing the data; selecting an optimal machine learning model; the average oxidation rate is obtained; and converting the average oxidation rate into oxidation weight gain. According to the method for predicting the oxidation weight increment of the gamma '-phase reinforced cobalt-based superalloy, the prediction model is constructed by depending on a component-heat treatment-average oxidation rate data set of the gamma'-phase reinforced cobalt-based superalloy and adopting a machine learning technology on the basis of a random forest algorithm, and finally the oxidation weight increment is obtained. According to the oxidation weight increment prediction method, the oxidation weight increment of the gamma'phase reinforced cobalt-based high-temperature alloy can be rapidly and accurately predicted within a large component and temperature range, and the method has high application value in component design and optimization of the series of alloys.

Description

Technical Field

[0001] The invention belongs to the technical field of cobalt-based high-temperature alloys, and in particular relates to a method for predicting oxidation weight gain of a γ′ phase-strengthened cobalt-based high-temperature alloy. Background Art

[0002] Cobalt-based high-temperature alloys play an important role in the aviation and energy industries due to their excellent heat corrosion resistance and thermal stability. They are used in key high-temperature components such as aircraft engines and gas turbines. In order to operate stably in harsh environments, these alloys need to have excellent density, high-temperature strength and oxidation resistance.

[0003] By rationally controlling the ratio of alloying elements (such as Al, Ti, Fe, Cr, Nb, Mo, W, Ta, Si, etc.) and optimizing process parameters, cobalt-based superalloys with excellent oxidation resistance can be prepared. In recent years, cobalt-based superalloys have gradually developed into a new type of alloy system. However, there is currently no mature method for predicting oxidation weight gain, which to a certain extent limits the rapid development and engineering application of this type of alloy.

[0004] Traditional predictions of superalloy oxidation weight gain mainly rely on classical theory and experimental determination. These methods often have many challenges, such as complex calculations of material behavior, estimation of multivariate parameters, and reliance on a large number of experiments. This not only increases costs, but also reduces the efficiency and accuracy of predictions.

[0005] In recent years, with the rapid development of artificial intelligence and computer science, machine learning technology has been introduced into materials science research and has provided a new solution for the prediction of oxidation weight gain of cobalt-based high-temperature alloys. Machine learning can efficiently process data sets and learn material properties through algorithms, significantly reducing experimental requirements, reducing R&D costs, and accelerating the development of new materials.

[0006] Therefore, a method for predicting the oxidation weight gain of γ′ phase-strengthened cobalt-based high-temperature alloys needs to be developed in this field. Through machine learning technology, a prediction model suitable for different temperature conditions is constructed to improve the prediction accuracy, reduce experimental costs, and promote the engineering application of such alloys. Summary of the invention

[0007] The purpose of the present invention is to provide a method for predicting the oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy. The method relies on a composition-heat treatment-average oxidation rate data set of a γ′ phase strengthened cobalt-based high-temperature alloy, and adopts machine learning technology to construct a prediction model based on a random forest algorithm, thereby realizing the prediction of the average oxidation rate of the series of alloys at different temperatures, and then converting the average oxidation rate into oxidation weight gain. The oxidation weight gain prediction method can make a fast and accurate prediction of the oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy within a larger composition and temperature range, and has a strong application value in the composition design and optimization of the series of alloys.

[0008] To achieve the above object, the present invention provides a method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy, comprising the following steps: Step S1, collecting a characteristic data set for predicting an average oxidation rate of a cobalt-based high-temperature alloy; Step S2, data preprocessing; Step S3, selecting the optimal machine learning model through a ten-fold cross validation method; Step S4: input the input parameters into the optimal machine learning model for prediction to obtain output parameters; Step S5: convert the predicted output parameters into oxidation weight gain.

[0009] Preferably, in step S1, the characteristic data is derived from existing literature, and relevant research is obtained by keyword retrieval, and the oxidation weight gain data of different alloys at various temperatures are extracted from the literature charts using Origin software; The characteristic data set includes alloy composition, heat treatment process parameters, cooling method, oxidation temperature, oxidation time, sample area and corresponding oxidation weight gain.

[0010] Preferably, in step S2, data preprocessing specifically includes visual cleaning of the data set and processing of oxidation weight gain by using an average oxidation rate formula; Step S21, data set visualization cleaning specifically includes: taking the oxidation time as the horizontal axis and the oxidation weight gain as the vertical axis, removing the missing and invalid data; Step S22: The average oxidation rate formula is specifically: ; In the formula, is the average oxidation rate value, in mg / (cm 2 h); Oxidation weight gain, unit is mg / cm 2 ; is the oxidation time, in h; Step S23: Take the natural logarithm ln .

[0011] Preferably, the data set has a size of more than 400 records, and the data set covers composition data of 11 alloy elements, more than 4 heat treatment process parameters and 2 cooling methods, and also includes oxidation weight gain at several temperatures and time points.

[0012] Preferably, the alloying elements include Co, Ni, Al, W, Ti, Ta, Cr, Mo, Nb, Si, and B; The heat treatment process parameters include but are not limited to solution temperature, solution time, aging temperature and aging time; the cooling methods include air cooling and water cooling, where air cooling is marked as "0" and water cooling is marked as "1".

[0013] Preferably, the content range of alloy elements is Co: 40.5-90at.%, Ni: 0-35at.%, Al: 8-10at.%, W: 0-3at.%, Ti: 0-10at.%, Ta: 0-2at.%, Cr: 0-10at.%, Mo: 0-5at.%, Si: 0-2at.%, B: 0-0.02at.%; The range of heat treatment process parameters is solution temperature 0~1350℃, solution time 0~24h, aging temperature 0~1000℃, aging time 0~168h; The parameter range of oxidation weight gain is: oxidation temperature 800~1150℃, oxidation time 0.8~600h; lnK is -8.2~1.7mg / (cm 2 h).

[0014] Preferably, in step S3, the machine learning algorithms considered when performing the ten-fold cross validation method include the nearest neighbor algorithm, random forest, gradient boosting algorithm and random forest algorithm; Among them, the parameter optimization range of the random forest algorithm is: number of trees: 20~200; depth: 2~20.

[0015] Preferably, in step S4, the input parameters of the optimal machine learning model are alloy composition, solution temperature, solution time, aging temperature, aging time, cooling method, oxidation temperature and oxidation time; wherein the output parameter is ln Predicted value.

[0016] Preferably, step S5 specifically comprises: a. First, calculate the predicted value. =e lnK ; In the formula, is the average oxidation rate value, in mg / (cm 2 h); b. Then use the average oxidation rate formula in reverse order. ; In the formula, is the average oxidation rate value, in mg / (cm 2 h); Oxidation weight gain, unit is mg / cm 2 ; is the oxidation time, in h; That is, the oxidation weight gain of the alloy at the oxidation temperature and oxidation time is obtained.

[0017] The present invention adopts the above-mentioned method for predicting oxidation weight gain of γ′ phase strengthened cobalt-based high-temperature alloy, and the beneficial effects are as follows: (1) In order to solve the problem that the data difference of oxidation weight gain is too large, resulting in high prediction error, the present invention calculates the average oxidation rate of the alloy The method is used and the natural logarithm is taken, thereby reducing the error and improving the prediction accuracy; (2) Taking into account the influence of factors such as the phase composition, distribution and size of the alloy on the oxidation resistance of the alloy, the present invention further introduces the regulation of heat treatment process parameters and cooling methods on the basis of optimizing the alloy composition, oxidation temperature and oxidation time; through this systematic optimization design, excellent error control and high-precision prediction model are finally achieved, so that the average oxidation rate performance of γ′ phase-strengthened cobalt-based high-temperature alloy can be effectively predicted, which significantly improves the reliability and practical value of the prediction; (3) The present invention obtains a method for predicting the average oxidation rate of γ′ phase-strengthened cobalt-based high-temperature alloys based on a random forest model by screening different machine learning algorithms. This method can quickly predict the average oxidation rate of the series of alloys under different compositions and process conditions at different temperatures, different cooling methods and different times, and finally obtain the oxidation weight gain. It not only has high prediction accuracy, but also has a wide range of applicable compositions and temperatures. (4) The specific implementation effect of the optimal model on the training set and the test set is significant. The results show that the accuracy value of the optimal model is R 2 It reaches 0.99, and the root mean square error is only 0.22, which has excellent prediction performance; therefore, the prediction method of the present invention has significant engineering application value in the composition design and process optimization of γ′ phase strengthened cobalt-based high-temperature alloys.

[0018] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1It is a flow chart of an embodiment of a method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to the present invention; Figure 2 This is a visualization and cleaning diagram of a data set of an embodiment of a method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to the present invention; Figure 3 This is a comparison chart between the actual and predicted oxidation weight gain values ​​of various alloys in an embodiment of a method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy of the present invention. DETAILED DESCRIPTION

[0020] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0021] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0022] Example like Figure 1 As shown, a method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy comprises the following steps: Step S1, collecting a characteristic data set for predicting the average oxidation rate of a γ′ phase strengthened cobalt high temperature alloy.

[0023] In this example, relevant literature was retrieved by searching keywords such as "high temperature alloy, oxidation", and the oxidation weight gain data of different alloys at various temperatures were extracted from the literature charts using Origin software. The complete data set includes 11 alloy components, more than 4 heat treatment process parameters, 2 cooling methods, oxidation temperature, oxidation time, sample area, and corresponding oxidation weight gain. A total of 604 data sets were obtained.

[0024] Step S2: data preprocessing.

[0025] like Figure 2 As shown, in this embodiment, the collected data set is visualized with oxidation time as the horizontal axis and oxidation weight gain as the vertical axis, and the lost and invalid data are removed; finally, 422 sets of data are retained as the training data set. The average oxidation rate formula (1) is used:

[0026] (1); Where: is the average oxidation rate value, in mg / (cm 2 h); Oxidation weight gain, unit is mg / cm 2 ; is the oxidation time, in h.

[0027] Convert oxidative weight gain into Then, Take the natural logarithm ln .

[0028] Step S3: Select the optimal machine learning model through the ten-fold cross validation method.

[0029] In this embodiment, the data set is randomly divided into a training set and a test set in a ratio of 8:2; the training set is used for model training, and the test set is used to evaluate the model accuracy. This embodiment adopts the ten-fold cross-validation technique to verify the generalization performance of the model and prevent overfitting in statistics. This technique randomly divides the data set into ten mutually exclusive subsets, selects nine of them as training sets each time, and the remaining one as a test set, and repeats ten times until all subsets are used as test sets. A variety of classic machine learning models are used to train the average oxidation rate training set of γ′ phase strengthened cobalt-based high-temperature alloys, and the average of each evaluation result is taken as a comprehensive evaluation index of model performance. By using the ten-fold cross-validation method to divide the data set, the root mean square error of multiple ten-fold cross-validations and the average value of the accuracy are obtained, and the accuracy value and root mean square error (RMSE) are used for evaluation. These two indicators can fully reflect the prediction accuracy and fit of the model. The specific formula is as follows: (2); (3); In formula (2), represents the total sample size of the dataset, represents the real value in the database, Represents the predicted value by the machine learning model.

[0030] The numerator in formula (3) represents the sum of squares of the difference between the predicted value and the true value, the denominator is the sum of squares of the difference between the true value and its mean, and R² represents the accuracy value.

[0031] Through the calculation of these two parts, the root mean square error can effectively reflect the prediction accuracy of the model. The value of the precision value (R²) is usually between 0 and 1. The closer the precision value is to 1, the better the model fitting effect and the stronger the prediction ability. Finally, the random forest model was selected with an RMSE of 0.22 and an R² of 0.99 to accurately predict the average oxidation rate of γ′ phase strengthened cobalt-based superalloy.

[0032] Step S4: input the input parameters into the optimal machine learning model for prediction to obtain output parameters.

[0033] In step S5, the predicted output parameter is converted into oxidation weight gain.

[0034] a. First, calculate the predicted value. =elnK (4); In the formula, is the average oxidation rate value, in mg / (cm 2 h); b. Then use the formula (1) in reverse order. (5); In the formula, is the average oxidation rate value, in mg / (cm 2 h); Oxidation weight gain, unit is mg / cm 2 ; is the oxidation time, in h; That is, the oxidation weight gain of the alloy at the oxidation temperature and oxidation time is obtained.

[0035] Four γ′ phase strengthened cobalt-based high-temperature alloy data were selected from outside the data set, including alloy composition, heat treatment process parameters, cooling method, oxidation temperature, and oxidation time as input parameters to the random forest model. The number of optimal parameter trees: 60, depth: 12, predicted that the alloy after heat treatment cooling, at oxidation temperature through oxidation time ln The four different alloy compositions and process parameters as well as the predicted and actual values ​​are shown in Table 1. Figure 3 As shown, the predicted ln obtained by the method of this embodiment Value and real ln The errors of the values ​​are 0.004, 0.006, 0.006, and 0.007 respectively.

[0036] Table 1 Four different alloy compositions and process parameters as well as predicted and actual values ;

[0037] The average oxidation rates of the four alloys can be obtained by the formula K1=1.08, K2=0.433, K3=1.53, K4=0.090. The predicted value of alloy oxidation weight gain was compared with the actual value, and the results are shown in Table 2.

[0038] Table 2 Comparison of predicted and actual weight gain of alloys due to oxidation ;

[0039] The errors between the predicted value of oxidation weight gain obtained by the method of this embodiment and the actual value of oxidation weight gain are 0.21, 0.05, 0.40, and 0.020, respectively.

[0040] Therefore, the present invention adopts the above-mentioned method for predicting the oxidation weight gain of γ′ phase strengthened cobalt-based high-temperature alloy. The method relies on the composition-heat treatment-average oxidation rate data set of γ′ phase strengthened cobalt-based high-temperature alloy, and adopts machine learning technology. A prediction model is constructed based on the random forest algorithm to realize the prediction of oxidation weight gain at different temperatures of this series of alloys; the oxidation weight gain prediction method can make fast and accurate predictions of the oxidation weight gain of γ′ phase strengthened cobalt-based high-temperature alloy within a larger composition and temperature range, and has strong application value in the composition design and optimization of this series of alloys.

[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based superalloy, characterized in that: The following steps are involved: Step S1, collecting a characteristic data set for predicting an average oxidation rate of a cobalt-based high-temperature alloy; Step S2, data preprocessing; Step S3, selecting the optimal machine learning model through a ten-fold cross validation method; Step S4: input the input parameters into the optimal machine learning model for prediction to obtain output parameters; Step S5: convert the predicted output parameters into oxidation weight gain.

2. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 1, characterized in that: In step S1, the characteristic data is derived from existing literature, and the research is obtained by keyword retrieval, and the oxidation weight gain data of different alloys at various temperatures are extracted from the literature charts using Origin software; The characteristic data set includes alloy composition, heat treatment process parameters, cooling method, oxidation temperature, oxidation time, sample area and oxidation weight gain.

3. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 1, characterized in that: In step S2, data preprocessing specifically includes visual cleaning of the data set and processing of oxidation weight gain by using the average oxidation rate formula; Step S21, data set visualization cleaning specifically includes: taking the oxidation time as the horizontal axis and the oxidation weight gain as the vertical axis, removing the missing and invalid data; Step S22: The average oxidation rate formula is specifically: ； In the formula, is the average oxidation rate value, in mg / (cm 2 h); Oxidation weight gain, unit: mg / cm 2 ; is the oxidation time, in h; Step S23: Take the natural logarithm ln .

4. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 3, characterized in that: The data set includes composition data of 11 alloying elements, 4 heat treatment process parameters, 2 cooling methods, and oxidation weight gain at several temperatures and time points.

5. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 4, characterized in that: Alloy elements include Co, Ni, Al, W, Ti, Ta, Cr, Mo, Nb, Si, and B; Heat treatment process parameters include but are not limited to solution temperature, solution time, aging temperature and aging time; cooling methods include air cooling and water cooling, where air cooling is marked as "0" and water cooling is marked as "1".

6. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 5, characterized in that: The content range of alloy elements is Co: 40.5~90at.%, Ni: 0~35at.%, Al: 8~10at.%, W: 0~3at.%, Ti: 0~10at.%, Ta: 0~2at.%, Cr: 0~10at.%, Mo: 0~5at.%, Si: 0~2at.%, B: 0~0.02at.%; The range of heat treatment process parameters is solution temperature 0~1350℃, solution time 0~24h, aging temperature 0~1000℃, aging time 0~168h; The parameter range of oxidation weight gain is: oxidation temperature 800~1150℃, oxidation time 0.8~600h; lnK is -8.2~1.7mg / (cm 2 h).

7. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 1, characterized in that: In step S3, the machine learning algorithms considered when performing the ten-fold cross validation method include the nearest neighbor algorithm, the random forest algorithm, the gradient boosting algorithm, and the random forest algorithm; Among them, the parameter optimization range of the random forest algorithm is: number of trees: 20~200; depth: 2~20.

8. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 1, characterized in that: In step S4, the input parameters of the optimal machine learning model are alloy composition, solution temperature, solution time, aging temperature, aging time, cooling method, oxidation temperature and oxidation time; the output parameter is ln Predicted value.

9. The method for predicting oxidation weight gain of a γ′ phase strengthened cobalt-based high-temperature alloy according to claim 3, characterized in that: Step S5 specifically includes: a. First, calculate the predicted value. =e lnK ； In the formula, is the average oxidation rate value, in mg / (cm 2 h); b. Then use the average oxidation rate formula in reverse order. ； In the formula, is the average oxidation rate value, in mg / (cm 2 h); Oxidation weight gain, unit: mg / cm 2 ; is the oxidation time, in h; That is, the oxidation weight gain of the alloy at the oxidation temperature and oxidation time is obtained.

Images