[2C01-04] Development of Calculation Code

[2C01] Prediction of Flow Behavior with Machine Learning Using Convolutional Neural Network

(1)Application to Cavity Flow

*Kenya Takiwaki¹, Aya Kitoh¹, Kei Meshima¹, Hideki Horie¹ (1. TOSHIBA ESS )

Keywords:machine learning , convolutional neural network , autoencoder, cavity flow

Recently, numerical analysis methods using machine learning has made rapid progress. Therefore, we investigated the applicability of machine learning to fluid analysis for the purpose of supporting the optimal design of the fluid equipment in power plants. In this study, we used a convolutional neural network autoencoder to predict the flow behavior of a two-dimensional square cavity flow. This is a method widely used in image recognition for extracting features of data and reconstructing. We used analysis results of a CFD code as training data, and predicted flow velocity distributions. As a result, characteristic secondary flows of the cavity flow appeared, and the mean absolute error was less than 1.5% of the flow velocity at the moving boundary.