Fire resistance evaluation through artificial intelligence - A case for timber structures

Abstract

Abstract With the ever-growing surge of new technologies, there seems to be an ongoing inertia towards integrating automation and cognition into various engineering applications. Despite a number of initiatives, and oddly enough of all civil engineering sub-disciplines, the structural fire engineering and fire safety community continues to embrace a classical stance to tackle the problem of fire. In support of growing demands to adopt performance-based solutions, this paper showcases the potential of integrating Artificial Intelligence (AI) as a unique technology to assess performance and fire resistance of structures. More specifically, this study sheds light on the proper use of AI to derive temperature-dependent material models for wood, together with simple expressions that can be used to trace thermo-structural response of timber elements/components (i.e. floor assemblies, beams, columns, and connections). These expressions comprehend the naturally complex temperature-induced physio-chemical changes to timber properties, including creep and charring, and hence do not require input of such properties nor special computing software. The outcome of this study clearly shows the merit of utilizing AI to modernize fire resistance evaluation given that the developed AI-models have high degree of perception (i.e. learn from past behaviors) and ability to improve their prediction capability through independent and unsupervised learning.