CN-122024937-A - Prediction model of mechanical properties of Al-Si alloy, construction method of prediction model and method for predicting mechanical properties of Al-Si alloy by using prediction model

CN122024937ACN 122024937 ACN122024937 ACN 122024937ACN-122024937-A

Abstract

The invention belongs to the technical field of metallurgical materials, and particularly relates to a prediction model of mechanical properties of an Al-Si alloy, a construction method thereof and a method for predicting the Al-Si alloy by using the model. According to the invention, by combining a machine learning method, the research and development period of the Al-Si alloy is shortened, and the optimal combination of the alloy components and the process is screened by using the machine learning model, so that the Al-Si alloy with good performances is easier to obtain, and a new thought is provided for developing and designing the Al-Si alloy with high mechanical performance based on machine learning. Moreover, in the process of establishing the data set, the maximum utilization of the published experimental data is achieved. In addition, by combining knowledge adding characteristics in the related material field, the prediction performance of the model is greatly improved, and the interpretation of the model is also improved.

Inventors

WANG DIANHUI
ZHAO YI
HU CHAOHAO
LU ZHAO
ZHONG YAN

Assignees

桂林电子科技大学

Dates

Publication Date: 20260512
Application Date: 20251010

Claims (10)

1. The construction method of the prediction model of the mechanical properties of the Al-Si alloy is characterized by comprising the following steps: (1) Carrying out data processing according to the known original data of the Al-Si alloy to construct an Al-Si alloy data set, wherein the original data comprises the element composition, the preparation process condition and the mechanical property data of each Al-Si alloy; (2) The method comprises the steps of carrying out normalization and standardization treatment on an Al-Si alloy data set, calculating the importance of each feature in an original data set by taking root mean square error as an evaluation index according to a cross-validation recursive feature elimination method of a self-adaptive enhancement algorithm, sorting the importance, screening the number and the variety of each feature, and dividing the screened data set into a training set and a testing set, wherein each feature is element composition, preparation process conditions and mechanical property data; (3) Training the model by different algorithms based on a training set, and improving the generalization capability of the model and obtaining a decision coefficient by five-fold cross validation training in the process of training the model; (4) And carrying out model fusion on the model with the highest prediction capability and the model with the strongest generalization capability to obtain an Al-Si alloy mechanical property prediction model, wherein the mechanical property comprises tensile strength, yield strength or elongation.
2. The method of claim 1, wherein the algorithm comprises one or more of a linear regression, a K-nearest neighbor algorithm, a decision tree, a random forest, a gradient lifting decision tree, a neural network, a lightweight gradient lifting machine, and an extreme gradient lifting algorithm.
3. The construction method according to claim 1, wherein the data processing includes deletion value filling, data screening, and cleaning performed sequentially; The filling of the missing value comprises the following specific steps: (1) According to the element composition in the initial data set, calculating physical characteristics which possibly influence the mechanical properties of the Al-Si alloy; (2) The element composition and the preparation process condition in the initial data set are used as characteristics, the tensile strength, the yield strength and the elongation are used as labels, and a random forest algorithm is used for filling the missing label data in the initial data set.
4. The method of claim 1, wherein the data is selected and cleaned to remove data having a Si content of less than 6wt.% and greater than 12wt.% from the raw data.
5. The method of claim 1, wherein the number of samples in the training set is 70-80% of the number of samples in the data set.
6. The construction method according to claim 1, wherein the model fusion is performed based on a Stacking method or a Blending method.
7. The method of constructing as claimed in claim 1, wherein said step of model fusion is: Taking a model with the highest prediction capability and a model with the strongest generalization capability as a base learner; and taking the prediction result of the base learner as the input parameter of the element learner and training the model to obtain the Al-Si alloy mechanical property prediction model.
8. A method for predicting an Al-Si alloy by using a prediction model of the mechanical properties of the Al-Si alloy comprises the following steps: Inputting the virtual database into the Al-Si alloy mechanical property prediction model constructed by the construction method according to any one of claims 1-7, and outputting the mechanical properties of the alloy under different element proportions; Obtaining the component composition and the preparation process condition of the target Al-Si alloy according to the predicted mechanical property result; the virtual database obtaining step comprises the following steps: (a) Determining the element types of the Al-Si alloy; (b) Processing the element types determined in the step (a) through data to generate a plurality of pieces of data with different element proportions; (c) And randomly matching the data with different element proportions with the preparation process conditions in the Al-Si alloy dataset to obtain a virtual database.
9. The method of claim 8, wherein the data processing of step (b) is to randomly generate a plurality of data of different element proportions using a dirichlet function in Python.
10. The computer readable storage medium is characterized by comprising an input module, an operation module and an output module, wherein the operation module is an Al-Si alloy mechanical property prediction model constructed by the construction method according to any one of claims 1-7, the input module is a virtual database, and the output module is the mechanical property of an alloy.

Description

Prediction model of mechanical properties of Al-Si alloy, construction method of prediction model and method for predicting mechanical properties of Al-Si alloy by using prediction model Technical Field The invention belongs to the technical field of metallurgical materials, and particularly relates to a prediction model of mechanical properties of an Al-Si alloy, a construction method thereof and a method for predicting the Al-Si alloy by using the model. Background Aluminum-silicon (Al-Si) alloys are widely used in the fields of automobile manufacturing, aerospace, electronic components and the like due to their excellent comprehensive properties, such as good casting properties, higher strength and hardness and the like. Meanwhile, the Al-Si alloy has low density and low cost, so that the Al-Si alloy also becomes an important material choice for realizing energy conservation and emission reduction. In recent decades, attempts have been made to improve the mechanical properties of al—si alloys in order to enhance the load-bearing capacity of the alloy material, extend its service life, and achieve a reduction in energy consumption. However, the limitations of mechanical properties of conventional Al-Si alloys are common, and a trade-off is required between strength and toughness. For example, a high silicon content can increase the proportion of eutectic Si, and after the eutectic Si is fibrillated, the casting fluidity can be improved, but pores can be introduced to reduce the strength, and the addition of Fe element can easily cause the generation of an alpha-Fe phase or a beta-Fe phase, wherein the beta-Fe phase can enhance the strength and improve the high temperature stability, but can easily cause the brittleness to increase to influence the elongation. In conclusion, how to make the Al-Si alloy have excellent comprehensive mechanical properties by precisely regulating and controlling the components of the alloy and the heat treatment process is one of the difficulties to be solved in the current material science field. However, the main process of the discovery of the new alloy in the prior art is alloy component design, smelting preparation, structural characterization and performance test, and then the alloy component and preparation process are determined according to the structural characterization and the performance test. Disclosure of Invention The invention aims to provide a prediction model of the mechanical property of Al-Si alloy, a construction method thereof and a method for predicting the Al-Si alloy by using the model, the method provided by the invention discovers that the new alloy can effectively save time and cost, improves the research and development efficiency, and simultaneously is easier to obtain the balance between the strength and the toughness of the Al-Si alloy, thereby obtaining the Al-Si alloy with good performances. In order to achieve the above object, the present invention provides the following technical solutions: The invention provides a construction method of a prediction model of mechanical properties of an Al-Si alloy, which comprises the following steps: (1) Carrying out data processing according to the known original data of the Al-Si alloy to construct an Al-Si alloy data set, wherein the original data comprises the element composition, the preparation process condition and the mechanical property data of each Al-Si alloy; (2) The method comprises the steps of carrying out normalization and standardization treatment on an Al-Si alloy data set, calculating the importance of each feature in an original data set by taking root mean square error as an evaluation index according to a cross-validation recursive feature elimination method of a self-adaptive enhancement algorithm, sorting the importance, screening the number and the variety of each feature, and dividing the screened data set into a training set and a testing set, wherein each feature is element composition, preparation process conditions and mechanical property data; (3) Training the model by different algorithms based on a training set, and improving the generalization capability of the model and obtaining a decision coefficient by five-fold cross validation training in the process of training the model; (4) And carrying out model fusion on the model with the highest prediction capability and the model with the strongest generalization capability to obtain an Al-Si alloy mechanical property prediction model, wherein the mechanical property comprises tensile strength, yield strength or elongation. Preferably, the algorithm comprises one or more of linear regression, K-nearest neighbor algorithm, decision tree, random forest, gradient lifting decision tree, neural network, lightweight gradient lifting machine and extreme gradient lifting algorithm. Preferably, the data processing comprises filling of missing values, data screening and cleaning which are sequentially carried out; The filling of the