Validation of a Model for Predicting Airtightness of Residential Units

Abstract

Airtightness plays a significant role in buildings energy efficiency. This paper describes validation of the new model for predicting airtightness values of residential units. This model utilizes a neural network in prediction of airtightness and is obtained based on in situ measurements at 58 units in the local Osijek area (Croatia) carried out during 2013. The model presents new approach to airtightness estimations by using 4 corrective factors associated with building envelope elements and their airtightness properties. The model was validated in local field conditions, but independent validation of the model in this paper was made on 5 residential buildings in the Republic of Serbia in order to determine its applicability on new data set outside local area of Osijek. The proposed model requires reduced amount of data for predicting airtightness values of residential units and therefore is faster and more economical than the actual field measurements. The proposed model could also be used for predicting airtightness values at the initial design phase already, as well as for planning systematic energy refurbishment of residential buildings in order to achieve adequate energy efficiency and appropriate thermal comfort in accordance with EU recommendations in this field.