Prediction of flow field and mass flow rate in a solar chimney at different heights using ANFIS technique

Abstract

Natural ventilation for buildings using solar chimney is increasingly attracting the attention of many researchers. Many techniques have been introduced to research on solar chimneys such as experimental, analytical, computational methods. Recently, with the development of computer technology, computational method, particularly, Computational Fluid Dynamics (CFD) becomes more common and widely applied in solar chimney, but this method still exists limitation. One of the main limitations is using much computational source. In this study, CFD was combined with Adaptive Neuro-Fuzzy Inference System (ANFIS) to prevail against this limitation when predicting flow field and mass flow rate in a chimney. In particular, the fluid flow and heat transfer in chimney were simulated with CFD to create dataset. Two ANFIS models were built, trained, and validated using dataset from CFD. After the training, ANFIS models can predict flow temperature, velocity and induced mass flow rate, respectively, with R-squared (R2) of 0.97, 0.997 and 0.9996 for training set, while 0.9715, 0.994 and 0.9996 for testing set; similarly, root mean squared error (RMSE) are 0.032, 1.703, 3.45x10−5 for training set, and 0.042, 1.713 and 2.95x10−5 for testing set. It is expected that the combination of CFD and ANFIS can estimate more different scenarios but using less computational time.