Prediction models for macro shrinkage of aluminum alloys based on machine learning algorithms

Abstract

Macro shrinkage percentage is one of the most important measures of the castability of alloys. To quickly and accurately predict the macro shrinkage of aluminum alloys is greatly meaningful before developing new components or new alloys. In this study, based on the dataset of macro shrinkage percentage of aluminum alloys obtained by our experiments, three algorithms, Support Vector Regression (SVR), General Regression Neural Network (GRNN) and Random Forest (RF), were adopted to construct the prediction models for macro shrinkage, respectively. K-fold cross-validation and analysis on stability and reliability of models indicate the RF model is the best one. Compared the prediction values of macro shrinkage percentage in binary, ternary and multicomponent aluminum alloys by RF model with the experiment data of this study and the data from literatures, the constructed model has been verified to have high prediction accuracy and reliability.