System Identification of ZnO Nanostructure Based On Physical Field Analysis and Deep Learning

Abstract

Abstract Zinc oxide nanowire is a widely used one-dimensional nanomaterial. The topological structure of nanowire determines its performance. At present, many researches have made on the growth conditions and structural parameters of zinc oxide nanowires through experimental methods. However, there is a lack of quantitative research on the correlation between macro physical fields and the parameter of structure. In this paper, we propose NPPN (Nanostructure-Parameter Prediction Network) based on deep learning algorithms to explore the correlation between physical fields and nanostructure parameters. The dataset is firstly created by experimental results and CFD simulation. We carry out large batch of preparation experiment of zinc oxide nanowires using the KQS automated machine based on the low-temperature two-step hydrothermal method. Meanwhile, we simulate the real physical field distribution of the reaction chamber through CFD software. The results show that the NPPN model has good predictive performance for the structural parameters of zinc oxide nanowires.