Deep learning method for predicting the mechanical properties of aluminum alloys with small data sets

Big data is usually needed for a deep learning method to predict the properties of materials, but, in practice, only limited data sets are available for engineering materials. In this study, we develop a deep neural network (DNN) to predict the mechanical properties of aluminum alloys from small data sets of chemical compositions and processing parameters. Through pre-training the DNN model to initialize a better set of parameters and tuning the model parameters, the present method can efficiently provide an accurate mapping between the mechanical properties and the composition and processing parameters of aluminum alloys. Compared with the support vector regression (SVR) and the shallow neural network (SNN), the present DNN method exhibits higher prediction accuracy and better generalization performance. It can be easily extended to other kinds of materials, and this work highlights the potential of the DNN method in data-driven material design. All original data and code that support the findings of this study are available from the corresponding authors upon request. • Fine-tunable deep neural network model for predicting the properties of alloys on small data sets. • This DNN method exhibits higher prediction accuracy and generalization performance. • Dependence of the strength of aluminum alloys on chemical compositions is examined.