An AI-based approach to thermal bridge analysis.

Buildings significantly contribute to climate change, accounting for approximately one-third of global energy consumption and a quarter of CO(2) emissions. Therefore, all actions aimed at increasing building energy efficiency are of great importance. This article proposes an original approach to analyzing thermal bridges in window-to-wall connections using artificial intelligence (AI). A fuzzy system (FS), employed as a universal approximator, proves particularly effective in this context. The FS, trained using data generated through conventional thermal bridge analysis conducted in the TRISCO program, offers interpretability that distinguishes it from other AI models. The FS was utilized to estimate linear heat transmittance coefficients, which quantify heat loss through thermal bridges. The proposed AI approach demonstrates excellent performance, generating precise linear heat transmittance coefficient values. Importantly, due to its ability to generalize knowledge, the trained system accurately determines the value of the Ψ coefficient for cases not included in the training data - those for which traditional analysis using the TRISCO program had not been previously performed. By leveraging this approach for thermal bridge analysis, it becomes possible to reduce the need for classical analyses, which often involve time-consuming calculations, expensive experiments, and extensive designer expertise in selecting optimal solution.