What question did this study set out to answer?

This research aims to explore the effectiveness of CNNs in predicting numerical fluid dynamics results under varying time conditions.

March 26, 2026Open Access

Investigation of numerical fluid analysis prediction using CNN

Key Points

This research aims to explore the effectiveness of CNNs in predicting numerical fluid dynamics results under varying time conditions.
Utilized CNNs to analyze flow fields as spatial information.
Compared predictions with analytical results under uniform time evolution.
Examined stability of CNN predictions with increased time evolution.
CNNs accurately predicted analytical results under consistent time evolution conditions.
The study suggests challenges may arise with increased time evolution during predictions.

Abstract

We have studied the prediction of numerical fluid dynamics results using AI techniques. In the past, we have used Convolutional LSTM to predict future spatial information from space-time information. We found that only CNNs with flow fields as spatial information can predict analytical results with high accuracy when the training data are under the same time evolution conditions. Therefore, we will examine whether the solution can be solved stably even when the time evolution of the numerical fluid dynamics analysis is greatly increased based on the results predicted using CNNs.

Investigation of numerical fluid analysis prediction using CNN

Key Points

Abstract

Cite This Study