Generalized temperature-dependent material models for compressive strength of masonry using fire tests, statistical methods and artificial intelligence

Masonry has superior fire resistance properties stemming from its inert characteristics, and slow degradation of mechanical properties. However, once exposed to fire conditions, masonry undergoes a series of physio-chemical changes. Such changes are often described via temperature-dependent material models. Despite calls for standardization of such models, there is a lack in such standardized models. As a result, available temperature-dependent material models vary across various fire codes and standards. In order to bridge this knowledge gap, this paper presents three methodologies, namely, regression-based, probabilistic-based, and the use of artificial neural (ANN) networks, to derive generalized temperature-dependent material models for masonry with a case study on the compressive strength property. Findings from this paper can be adopted to establish updated temperature-dependent material models of fire design and analysis of masonry structures.