[2C01-04] Development of Calculation Code

[2C02] Prediction of Flow Behavior with Machine Learning Using Convolutional Neural Network

(2)Application to Backward-facing Step Flow

*Aya KITOH¹, Kei MAESHIMA¹, Kenya TAKIWAKI¹, Hideki HORIE¹ (1. TOSHIBA ESS)

Keywords:machine learning, deep learining, CNN, CFD, backward-facing step

Recently, research applying machine learning and deep learning to engineering numerical analysis, such as computational fluid dynamics (CFD), has made rapid progress. In this study, we investigated the applicability of machine learning to CFD, for the purpose of supporting the design and development of fluid equipment for plants, such as nuclear power. This time, we evaluated the flow behavior using a CNN autoencoder. Specifically, targeting the steady flow around a backward-facing step, 100 results obtained with CFD on the inlet Reynolds number (Re) of 10-1000 were used as training data. The flow-velocity distributions of the Re that were not used for the training were predicted, and we obtained the result with the mean squared less than 1%.