An approach for a novel multiscale partitioned framework for predicting collapse of buildings under fire loads

The structural integrity and load-bearing capacity of burning buildings can be significantly reduced due to high thermal load.This causes a problem for e. g. rescue workers, as they have to ensure the rescue of people in a burning building even if there is an increased risk of collapse.Predicting safe routes in and out of a burning building is a difficult task that takes resources and expertise.Alternatively, escape routes can be identified using simulations, which require a multidisciplinary and multiscale approach.These simulations may either run in real-time or need to be coupled with artificial neural networks in order to train them such that these neural networks suggest safe escape routes in real-time in case of a fire incident.Computational-fluid-dynamic simulations of the fire provide information about both the temperature and the smoke distribution in the single rooms of a building.Micro-scale simulations describe local material behaviour such as spalling and cracks.Structural-scale analysis provides information about the global structural behaviour of the building including a possible progressive collapse.Nevertheless, a coupling of these multi-physical simulations is needed to identify safe escape routes in a burning building.However, it is not only necessary to couple across scales but also across software platforms, as there is -to the best knowledge of the authors -no simulation framework available that can calculate the stated effects on multiple scales with good accuracy.Consequently, studying these fire-structure interactions necessitates an accurate computational simulation framework that integrates the effects of different physical phenomena.The objective of this contribution is to provide a computational framework for fire-structure simulations to elucidate the interplay between damage of concrete and progressive collapse.In the proposed computational setup, two software packages are used to simulate material damage and the resulting collapse of an exemplary structure under fire.A micro-scale model uses a FEniCs-based solver for the concrete material and a structural-scale model uses AB A Q U S for the (progressive) collapse simulation.These solvers exchange information on temperature, stress, and changes in geometry across scales and software through an open-source coupling library, preCICE.The proposed model is illustrated by a numerical experiment that forecasts the damage leading to progressive collapse.