Modeling of hygrothermal behavior for green facade's concrete wall exposed to nordic climate using artificial intelligence and global sensitivity analysis

Abstract Green facades are one of the most promising natural-based solutions for buildings. Notwithstanding, in regions with variating weather such as the northern hemisphere, these can be counterproductive for the structures due to humidity retention. For these reasons, this work presents the development of an Artificial Neural Network (ANN) model to estimate the hygrothermal behavior inside a concrete wall protected by a second foliage skin. The database used for model formation was obtained through measurements made in an Accelerated Weathering Laboratory (AWL) to emulate the Nordic climatic conditions for a typical year. The ANN-hygrothermal model was trained in function of the parameters: environment relative humidity, ambient temperature, microclimate's relative humidity, microclimate's temperature, and the separation distance between the vegetation and the wall. The statistical results of the model demonstrated successful adaptability and great generalization capacity for both internal temperature (R2 = 99.98% for training and R2 = 99.95% for testing) and internal humidity (R2 = 99.16% for training and R2 = 99.17% for testing). Additionally, a sensitivity analysis was implemented, showing that the most influential variable in the estimation of both hygrothermal parameters is the ambient temperature and that the separation distance has a significant impact on the humidity produced inside the wall. Finally, the presented computational approach can be implemented in non-invasive monitoring systems or as a complementary tool in studies of concrete degradation due to humidity.