Numerical and surrogate modeling of drying processes in building envelopes under variable climatic conditions

Variable climate conditions, including increasingly frequent extreme weather events such as floods and prolonged rainfall, pose a significant challenge to the durability and functionality of building materials. This study analyzes the drying time of multilayer wall partitions subjected to fluctuating moisture loads and different initial moisture levels. A series of hygrothermal simulations was conducted using WUFI software, which solves coupled nonlinear differential equations describing heat and moisture transport in porous materials. The simulations were carried out for various wall configurations commonly used in construction, including autoclaved aerated concrete, ceramic blocks, silicate blocks, and concrete units, using climatic data for Warsaw. Initial moisture levels ranged from air-dry to fully saturated conditions. The results indicate that the drying process strongly depends on the type of material; aerated concrete and silicate blocks showed faster drying rates, while some layers, such as gypsum boards and adhesives, demonstrated significant moisture accumulation during the early stages. Based on these results, a surrogate model was developed using stochastic modeling techniques and principal component analysis (PCA) to compress the simulation data and predict drying times. This model enables rapid estimation of drying periods under various exposure scenarios, supporting improved design and risk assessment after water damage to building envelopes. • Drying processes of building envelopes were analyzed using numerical modeling in WUFI 2D with real climatic data for Warsaw. • Long-term computer simulations (3 years, multiple initial moisture levels) were performed to assess moisture variation in different wall layers. • Four wall materials were compared: autoclaved aerated concrete (AAC), silicate blocks, ceramic blocks, and regular concrete. • The fastest drying was observed in AAC and silicate blocks, while ceramic and concrete walls exhibited much slower drying. • A surrogate model based on PCA and Gaussian Process Regression was developed to replicate full-scale simulations at a fraction of computational cost. • The surrogate enables rapid drying time prediction and mold risk assessment with ∼96% accuracy, supporting engineering practice and post-flood renovation planning.