Application of ANN to explore the potential use of natural ventilation in buildings in Turkey

Abstract

Indoor natural ventilation provides both the circulation of clear air and the decrease of indoor temperature, especially, during hot summer days. In addition to openings, building dimensions and position play a significant role to obtain a uniform indoor air velocity distribution. In this study, the potential use of natural ventilation as a passive cooling system in new building designs in Kayseri, a midsize city in Turkey, was investigated. First, indoor air velocity distributions with respect to changing wind direction, magnitude and door openings were simulated by the FLUENT package program, which employs finite element methods. Using the simulated data an artificial neural network (ANN) model was developed to predict indoor average and maximum air velocities. The simulations produced by FLUENT show that the average indoor air velocity is generally below 1.0 m/s for the local prevailing wind directions. The simulations results suggest that, in addition to the orientation of buildings in accordance with prevailing wind directions, a proper indoor design of buildings in the area can significantly increase the capability of air ventilation during warm summer days. It was found that a high correlation exists between the simulated and the ANN predicted data indicating a successful learning by the proposed ANN model. Overall, the evaluation of the network results indicated that the ANN approach can be utilized as an efficient tool for learning, training and predicting indoor air velocity distributions for natural ventilation.