Natural ventilation and artificial neural networks.

This study presents a new method of interior air motion assessment using artificial neural networks. The air motion inside a building depends not only on the external wind velocity, but also to a great extent on most of architectural parameters such as position and orientation of building, size and configuration of windows, roof geometry, whether the building is stilted or not, etc .. The difficulty to evaluate the interior velocity coefficient, a non-dimensional parameter that is the measure of relative strength of the interior air movement, if we would take into account a number of architectural parameters; this encouraged us to use this approach. After presenting the general setting of our work, we introduce the neural networks in describing their main properties and the methods of their implementation. We have applied these ideas to our study and presented the initial obtained results. The utilization of the neural networks as a model-free predictor is a way of interesting investigation which facilitates designers or architects to take into account a number of influential parameters in natural ventilation investigating. Moreover,  this allows to assess indoor airflow pattern without doing a costly experiment or running an expensive and complicated flow field simulation code.