Design and performance prediction of high entropy alloy based on machine learning
Background: High entropy alloy (HEA) is a kind of alloy composed of five or more elements with approximately equal molar ratio, which has excellent properties such as high mixing entropy, high thermal stability, high strength and high toughness. It is a new kind of material with broad application prospects. However, the design of high entropy alloys is faced with huge composition space and complex performance prediction problems, and the traditional experimental methods are difficult to meet the needs of high efficiency, accuracy and intelligence. Therefore, using machine learning (ML) technology, combining theoretical calculation and experimental data, to establish the design and performance prediction model of high entropy alloy is an effective solution.
The purpose of this project is to develop a machine learning-based design and performance prediction method for high-entropy alloys to achieve rapid screening, optimization and evaluation of high-entropy alloys. This project will use a variety of machine learning algorithms, such as support vector machine, neural network, random forest, etc., combined with a variety of feature selection and dimensionality reduction technologies, such as principal component analysis, genetic algorithm, Shapley value, etc., to build a composition-performance relationship model of high entropy alloy, and use cross-validation, grid search, Bayesian optimization and other methods to train and optimize the model. In this paper, the hardness, strength, toughness and corrosion resistance are taken as the target properties, and Al-Co-Cr-Cu-Fe-Ni is taken as the target system to design and predict the properties of high-entropy alloys, and compare and verify with the experimental data.
Project content:
Data collection and pre-processing: The composition and performance data of high-entropy alloys of the target system were collected from sources such as reference 4, and pre-processing operations such as data cleaning, normalization and balancing were carried out to obtain data sets that could be used for machine learning.
Feature extraction and selection: A variety of physical, chemical and statistical characteristics, such as atomic radius, electronegativity, mixing enthalpy, mixing entropy, band gap, atomic size difference, etc. were calculated according to the composition information of the high entropy alloy, and principal component analysis (PCA), genetic algorithm (GA), Shapley value (SHAP) and other methods were used for feature selection and dimensionality reduction. Screen out the features that have an important impact on performance prediction.
Model construction and optimization: A variety of machine learning algorithms, such as support vector machine (SVM), neural network (NN), random forest (RF), etc., were used to build a composition-performance relationship model of high-entropy alloy based on the selected features and target performance, and the model parameters were adjusted and optimized by cross-validation (CV), grid search (GS), Bayesian optimization (BO) and other methods. Improve model accuracy and generalization ability.
Model evaluation and validation: A variety of evaluation indicators, such as mean square error (MSE), determination coefficient (R2), and mean absolute percentage error (MAPE), were used to evaluate the prediction effect of the model on the training set and test set, and to compare and verify the model with the experimental data to analyze the advantages and disadvantages of the model and the scope of application.
High-entropy alloy design and prediction: Using the trained model, the high-entropy alloy of the target system is quickly screened, optimized and evaluated, and the high-entropy alloy components with optimal or near optimal properties are found, and their performance values are predicted to provide guidance and reference for experimental design.
Topic innovation:
In this paper, a kind of efficient, accurate and intelligent design and performance prediction method of high entropy alloy is established by using a variety of machine learning algorithms, feature selection and dimensionality reduction techniques, model optimization and evaluation methods, which provides a new idea and tool for the discovery and development of high entropy alloy.
This subject takes hardness, strength, toughness, corrosion resistance as the target properties, and Al-Co-Cr-Cu-Fe-Ni as the target system, covering a variety of important properties and systems of high entropy alloys, with strong practicability and universality.
In this paper, the interpretability of the model is improved by using Shapley value and other methods, and the internal relationship and influencing factors between the composition and properties of high entropy alloys are revealed, which provides a basis and guidance for the theoretical study of high entropy alloys.
| Kind reminder: Suzhou Beike Nano supply products are only used for scientific research and cannot be used for the human body. The specifications and performance of different batches of products may vary. Some literature case images on the website are sourced from the internet and are for reference only. Please focus on physical objects. If there is any infringement, please contact us to delete them immediately. |
Previous article： Preparation and charac
Next article： Life Sciences - Micro-
