Tajima, M.; Sawachi, T.
Bibliographic info:
28th AIVC and 2nd Palenc Conference " Building Low Energy Cooling and Ventilation Technologies in the 21st Century", Crete, Greece, 27-29 September 2007

The objective of this paper is to evaluate the usefulness and the accuracy of an artificial neural network (ANN) as a prediction tool of the wind pressure coefficient (Cp). The ANN is applied to predict the Cp for rectangularbuildings. The Cp values obtained by wind tunnel experiments are approximated by using a Cascade-CorrelationLearning Network model to make a prediction, and the performance of three kinds of residential ventilationsystems depending on the wind effect is evaluatedby using both of the predicted Cp and the Cp experimentallyobtained. The results of the approximation for the rectangular buildings of three different heights suggest that the ANN has a promising potential as a prediction tool for the Cp with sufficient accuracy. The performance of the ventilation systems is evaluated of their total ventilation rates and fulfilment of ventilation requirements in multiple rooms represented by "OverallSupply Rate Fulfilment" values. In the performance evaluation, similar result is obtained even when the predictedCp is used in the network simulation.