A model was developed to predict fully guarded blower door results, using unguarded test data and the building features of apartment units. Data comprised of 202 units from four climate zones, 15 multi-family buildings and 36 possible predictor variables. The model was trained with 121 units and the remainder was used to validate the model. Four variables were selected as most significant based on variables’ availability in data, stepwise regression analysis, ease of obtaining variables and findings from previous research. The final model uses the variables: Ductwork Location (conditioned space, unconditioned space, none), RESATSA, Exposed Floor Area and Age. The RMSE of the model was 0.14 and the RMSE of the predicted ratio was 0.12. The result of this research is certainly limited by the training data used. Thus, the rigor of the predictive model could be improved with the addition of more buildings with different types of construction, different climate regions, etc. As such additional work is needed to validate and improve the model with more data.
Citation: ASHRAE Papers CD: 2014 ASHRAE Winter Conference, New York, NY
Product Details
- Published:
- 2014
- Number of Pages:
- 8
- File Size:
- 1 file , 2 MB
- Product Code(s):
- D-NY-14-C040
