Project background:
High entropy alloy is a kind of alloy composed of five or more elements with approximately equal molar ratio. It has excellent physical and chemical properties such as high mixing entropy, high thermal stability, high strength, high toughness and high corrosion resistance. It is a new kind of material with broad application prospect. However, the design and optimization of high-entropy alloys face great challenges because the composition space is so large that traditional experimental methods are difficult to cover all possible combinations. Therefore, it is an effective and efficient method to predict and screen the composition and properties of high entropy alloys by using machine learning technology combined with theoretical calculation and experimental data.
Project information: This project aims to use machine learning technology to establish a high-entropy alloy design and performance prediction model based on multi-omics data (such as structure, mechanics, thermodynamics, dynamics, etc.), to provide guidance and support for the discovery and development of high-entropy alloys. This project will use deep neural network, support vector machine, random forest and other methods to obtain multiple omics data from public databases or cooperative laboratories, and carry out data cleaning, feature extraction, feature selection, model training, model evaluation and other steps to build and optimize the prediction model, and verify and apply it.
Project content:
1. Data acquisition and preprocessing: Obtain multiple omics data from public databases or cooperative laboratories, and conduct data cleaning, normalization, dimensionalization and other preprocessing steps to facilitate subsequent analysis.
2. Feature extraction and selection: Use deep neural networks or other methods to extract features related to the composition and performance of high-entropy alloys from multi-omics data, and carry out feature selection and fusion to form a high-dimensional feature space.
3. Model training and evaluation: Support vector machine, random forest or other methods are used to build regression or classification prediction models based on feature space and performance data, and cross-validation, grid search and other methods are used for model training and parameter adjustment, and correlation coefficient, mean square error, accuracy, recall rate and other indicators are used for model evaluation and comparison with existing methods.
4. Model verification and application: Independent test sets or external data sets were used to verify the prediction model, and multi-objective optimization and screening were carried out using Pareto frontier analysis and other methods to obtain the candidate composition of high-entropy alloy with the best combination of properties, and experimental preparation and characterization were carried out to verify the accuracy and reliability of the model. And explore the applicability and universality of the model in different types, different targets and different conditions.
Topic innovation:
1. This topic will use machine learning technology and multi-omics data to establish a comprehensive and accurate design and performance prediction model for high-entropy alloys, providing new ideas and methods for the discovery and development of high-entropy alloys.
2. This topic will use advanced algorithms such as deep neural networks to learn effective features and patterns from large-scale multi-omics data to improve the performance and accuracy of prediction models.
3. This topic will systematically train, evaluate and verify the prediction model, and compare it with existing methods to show the advantages and limitations of the model, and explore the application and promotion potential of the model in different high-entropy alloy scenarios.