Comprehensive in-situ assessment of glazing systems: thermal properties, comfort impacts, and machine learning-based predictive modelling

In-situ methods have the potential to reliably evaluate building façade components, including glazing systems. These methods have the potential to assess key glazing properties such as thermal transmittance (U-value) and solar heat gain coefficient (g-value), as well as aspects like thermal comfort near the glazing and the impact of dynamic outdoor conditions on glazing performance. This study employs an extensive in-situ evaluation approach, utilizing an average-based strategy to determine the U-value and g-value of a double-glazed unit while analyzing the influence of solar radiation on occupants’ thermal comfort near glazing. Additionally, the study explores the potential of machine learning to predict variations in glazing surface temperatures across seasons, based on a relatively short measurement period. Results indicate that the standard deviations for the measured U-value and g-value across seasons range from approximately 8 % to 25 % and 2 % to 10 %, respectively. Solar radiation significantly affected thermal comfort near the glazing, increasing the Predicted Percentage Dissatisfied (PPD) up to threefold in summer, causing discomfort, while reducing it by half in winter, improving comfort. Although machine learning predictions correlated strongly with nighttime measurements, discrepancies emerged during the day due to the highly dynamic nature of solar radiation, making daytime predictions more challenging than nighttime ones. Nonetheless, general variation patterns were reasonably captured. The study concludes by proposing a comprehensive approach that integrates reliable laboratory methods with in-situ evaluations to more effectively test the reliability of the adopted method.