On the impact of regression technique to airtightness measurements uncertainties

摘要

The Ordinary Least Square (OLS) method is often used in the estimation of buildings' airtightness. However, recent research demonstrated that this regression technique neglects a significant part of uncertainty. This paper investigates two alternative methods: Weighted Line of Organic Correlation (WLOC) and Iterative Weighted Least Square (IWLS). A series of 30 repeated tests were performed on a newly-constructed apartment in Brussels (Belgium). Real uncertainty and its estimation were compared for the three regression methods. In-situ observations showed that the statistical significance of differences between regression techniques depends on the considered pressure difference. Theoretical calculations showed that uncertainties estimated with IWLS and WLOC were higher than OLS. Lastly, comparing calculations and in-situ observations confirmed that IWLS and WLOC provide a better uncertainty estimation than OLS. Therefore, WLOC or IWLS should be preferred to OLS, especially in the presence of wind. WLOC is expected to be more robust than IWLS since it does not imply an iterative procedure. Along with ongoing work investigating other sources of errors, these findings contribute to a deeper understanding of uncertainty in fan pressurization tests. (C) 2020 Elsevier B.V. All rights reserved.