Fawzy, Diaa E.Arslan, Guvenc2023-06-162023-06-1620151877-0428https://doi.org/10.1016/j.sbspro.2015.06.179https://hdl.handle.net/20.500.14365/1419World Conference on Technology, Innovation and Entrepreneurship -- MAY 28-30, 2015 -- Istanbul, TURKEYThe current work is an attempt to predict building reactions to big earthquakes using real data collected from surveys carried out after the occurrence of earthquakes. With the development of building damage functions for big earthquakes in Turkey one can predict the damage levels as a function of earthquakes' intensity and the building parameters. Our model is based on neural networks techniques which allow for the non-linear correlations to be taken into account. We analyse data collected for damaged buildings after the following three big earthquakes: Afyon (2002; Mw - 6.0), Bingol (2003; Mw - 6.4) and Duzce (1999; Mw - 7.2). The current model includes some of the main important factors affecting the health of any structure, namely, age, number of stories, floor areas, and the column areas. Our method of damage prediction is based on several earthquakes and buildings with different damage levels. The obtained results show that there is a strong correlation between the strength of the earthquake, the basic building parameters and the damage level. The obtained building damage function is essential for future plans and regulations for new constructions and can be considered as an essential module for hazards mitigation systems. (C) 2015 The Authors. Published by Elsevier Ltd.eninfo:eu-repo/semantics/openAccessearthquakesneural networksstructural health monitoringestimation methodsConference Object10.1016/j.sbspro.2015.06.179