Evaluation of the applicability of the extended average method for non-destructive in-situ measurement of thermal transmittance in highly insulated walls

Accurate in-situ evaluation of the thermal transmittance of highly insulated walls is essential for assessing the energy performance of buildings. The heat flow meter (HFM) method specified in ISO 9869-1 is widely used, but its accuracy is often limited under varying environmental conditions or in high-insulation assemblies. In this study, the extended average method (EXAM) was further developed by integrating data-driven virtual sensing to enable non-destructive U-value estimation, specifically focusing on an internally insulated concrete wall. Multiple linear regression, artificial neural networks, and support vector machines (SVM) were developed to estimate the heat fluxes and temperatures of the embedded walls, and the EXAM results were compared with those of the average and dynamic HFM methods. The virtual-sensing-based EXAM exhibited closer agreement with the ISO 6946 reference value compared with the HFM methods, with the greatest improvement for large temperature differences between indoor and outdoor areas. It also increased the likelihood of meeting the ISO 9869-1 convergence criteria. Although some variations were observed between the individual models, the SVM-based EXAM consistently provided stable estimation performance and closer agreement with the reference value. These results demonstrate the potential of EXAM as a reliable damage-free approach for U-value estimation, serving as a foundational step for broader applications across various wall types.