AI-enhanced manufacturing: a common data-driven framework for industrial applications
Dr Federico Zipoli, IBM Researcher
We present a data-driven approach for formulations of novel materials via autoencoder-based models.1 Inspired by the works by Kingma2 and Bombarelli,3 we make use of deep-learning techniques to search for important correlations and patterns in the underlying data to improve existing products and design new ones. Starting from data that can consist of various inputs like compositions, processes, or even ageing conditions, and as outputs, product properties, we encoded the input into three latent representations utilizing encoder-decoder neural-network structures. From these trained latent space vectors, the property can be predicted by a separate feed-forward neural network. The scheme has been successfully applied to various material case studies ranging from polymers, epoxy resin, to different alloys. As an example of the broad potential use of such an algorithm, we present an application to the prediction of crystalline structures and of phase diagrams of ceramic materials using commercially available databases.
References:
1) T. Gaudin, O. Schilter, F. Zipoli and T. Laino, https://ercim-news.ercim.eu/en122/special/advanced-data-driven-manufacturing.
2) D. P. Kingma and M. Welling, Auto-encoding Variational Bayes, ICRL 2014, https://arxiv.org/abs/1312.6114
3) R. Gómez-Bombarelli, J. N. Wei, D. Duvenaud, J. M. Hernández-Lobato, B. Sánchez-Lengeling, D. Sheberla, J. Aguilera-Iparraguirre, T. D. Hirzel, R. P. Adams and A. Aspuru-Guzik, ACS Central Science, 2018, 4, 268–276.