Open AccessNeural network for indoor airflow prediction with CFD database
Author(s) -
Qi Zhou,
Ryozo Ooka
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/2069/1/012154
Subject(s) - computational fluid dynamics , airflow , cfd in buildings , artificial neural network , computer science , indoor air quality , energy consumption , simulation , efficient energy use , building energy simulation , thermal comfort , thermal , environmental science , marine engineering , artificial intelligence , engineering , meteorology , energy performance , aerospace engineering , mechanical engineering , physics , electrical engineering , environmental engineering
Energy efficiency and indoor thermal comfort are both important in built environment, making it necessary to simultaneously take into consideration of the two aspects, building energy performance and indoor environmental quality, at the design stage. Coupled simulation between building energy simulation (BES) and computational fluid dynamics (CFD) enables providing each other complementary information with regard to building energy performance and detailed indoor environment conditions; however, the main drawback of CFD in computational cost limits its application. Neural networks (NNs) are considered as promising alternatives for CFD due to their advanced modelling abilities and high-speed computational powers. This research aims to confirm the feasibility of NN for indoor airflow prediction, which extends previous studies from two-dimensional to three-dimensional indoor space for more realistic conditions. The NN receives boundary conditions as input and outputs corresponding velocity and temperature distributions. Comparisons were made between NN predictions and CFD simulations regarding accuracy and time consumption on testing cases. The results show that the NN reproduces indoor airflow and thermal distributions with relative errors less than 12%. Time consumption for predicting the testing cases is reduced by 80% with the NN. The feasibility of NN for fast and accurate indoor airflow prediction is confirmed.