Login
Purpose of the work
The purpose of this work is to evaluate and compare various regression techniques—specifically Ordinary Least Squares (OLS), Weighted Least Squares (WLS), and Weighted Line of Organic Correlation (WLOC)—in the context of building airtightness measurements as prescribed by ISO 9972. The goal is to improve the accuracy and reliability of uncertainty estimates in airtightness tests, particularly under varying environmental conditions such as different wind speeds. This research aims to inform enhancements in building airtightness testing standards to ensure more accurate measurements and better compliance with regulatory requirements.
Method of approach
The study employs a comprehensive dataset of over 6,000 blower door tests conducted across six test houses. The data is filtered according to ISO 9972 requirements, and the selected regression techniques are applied to model the relationship between pressure differences and airflow rates. The accuracy of each method is assessed by comparing predicted values and confidence intervals against reference values determined under low-wind conditions. The study also includes uncertainty estimation using procedures outlined in the GUM and validates the methods through comparison with observed data variability under different environmental conditions.
Content of the contribution
The core of the analysis involves applying three regression methods. Each method is examined for its effectiveness in predicting airflow rates at two key pressure points: 50 Pa, commonly used as a standard reference in building codes, and 4 Pa, which is more reflective of real-world building performance under natural conditions.
The study conducts a comparison of the regression methods by calculating and analyzing the uncertainty and confidence intervals for each technique. Specifically, the analysis evaluates how well each method predicts the airflow rates under different wind speeds, comparing the predicted results to reference values obtained under low-wind conditions. The methods are further assessed for their ability to produce reliable 95% confidence intervals, which indicate the range within which the true airflow rate is expected to fall.
The analysis also includes an investigation into the impact of systematic errors, such as those introduced by fluctuating wind speeds, on the accuracy of the regression methods. By comparing the predicted airflow rates and confidence intervals across different environmental conditions, the study identifies the strengths and limitations of each regression technique.
Results and assessment of their significance
The results indicate that both WLS and WLOC methods outperform the traditional OLS method, particularly in conditions with higher wind speeds. WLS shows the highest overall coverage of the 95% confidence intervals, making it a reliable choice for estimating uncertainty, though it tends to be overly conservative at lower pressures. WLOC, especially when simplified (WLOC_2), provides a good balance between accuracy and reliability, suggesting it could be a valuable addition to current airtightness testing standards. The significance of these findings lies in their potential to improve the precision of airtightness tests, which is crucial for energy efficiency, indoor air quality, and regulatory compliance in building construction and renovation.
Conclusions
This study demonstrates that alternative regression techniques, such as WLS and WLOC, can enhance the accuracy and reliability of building airtightness measurements compared to the conventional OLS method recommended by ISO 9972. The findings support the incorporation of these methods into standardized testing procedures, offering better uncertainty estimation and more accurate predictions of airflow rates under varying environmental conditions. By improving airtightness testing standards, this research contributes to the broader goals of energy efficiency, sustainability, and occupant comfort in the built environment.
For further information please contact Benedikt Kölsch at: benedikt.koelsch@dlr.de