Enhancing CFD computational efficiency using hybrid data-driven and physics-based modeling
Aliasghar Azma, Yakun Liu
Abstract
Computational Fluid Dynamics (CFD) is commonly used to simulate the transport of heat in closed spaces. The resulting airflow and temperature predictions facilitate improved designs of Heating, Ventilation, and Air Conditioning (HVAC) systems. However, CFD is highly expensive to apply to large domains. This paper presents a novel approach that is a hybridization of artificial intelligence (AI) with CFD modeling, which improves computational speed and predictive accuracy. Specifically, CFD data from a forced air-conditioned room is used to train an Adaptive Network-based Fuzzy Inference System (ANFIS) with temperature taken to be the dependent variable. The trained ANFIS predicts the temperature distribution on a high-resolution mesh using partial CFD data without need for additional numerical modelling. Results for transient hot air inflow to an idealized ‘room’ demonstrate that ANFIS is a very useful adjunct to the CFD method with high accuracy achieved using coarse-grid CFD data. The proposed AI-CFD hybrid framework should enable fast, efficient HVAC system designs that are more sustainable through reducing energy consumption and computational overhead. Moreover, the framework could facilitate real-time energy monitoring of buildings.