Numerical modelling of air flows in buildings and design of a data base of experiments

During the recent years a lot of models describing air flows in buildings have been designed by Researchers. These models, which are based more or less on simplified approaches coming from fluid dynamics, share a high complexity level because of complex geornetrical structure and. complex components of buildíngs. This complexity is the source of bad, experimental fitting, most of all because scale models don't correspond to the needs for fitting. 
From this background, the International Energy Agency has organized a search all over the world to collect information on projects which try to build data bases based on experiments. 
We have started a large project with this goal. A flat at scale one with six rooms and an entrance hall has been equipped with. measurement systems. We are able to control and generate air flows in the flat. Air flow characteristics of components; (walls, doors ... ) are known. So we are able using a gas tracking technique to determine air flows between rooms. In the same time we simulate; air flows characteristics in the flat with. a numerical code : STAR-CD1 . 
This has led us to a comparison of numerical and experimental results. We have found differences and failings of the code. These aspects are developed in the paper. They are the starting point of a discussion on CFD capabilities in the field of air flows simulation in buildings and especially when singularities ( cracks or lacks of windows ... ) are involved. We conclude by a look over local models which may lead to a better knowledge, and of course, simulation of the singularities.