Data-driven-based models for seismic damage assessment of buildings : using machine learning at a regional scale and in-situ structural monitoring at building scale

Overview & Synopsis

Decision-makers and stakeholders must have quick assessment of potential damage from earthquakes and its distribution to plan for successful emergency response. There are several methods for estimating building damage from earthquakes, but they involve many assumptions associated to a large set of parameters that controls the building response during earthquake, that can add uncertainty. Using real-world data from full-scale observations can help better understand the physical processes involved during seismic loading in building and identify sources of uncertainty in damage estimation. There is now access to a large amount of information describing the real estate building portfolios, collected through post-earthquake damage surveys at regional level and in-situ measurements in building. The objective of this study is to test the effectiveness and relevance of supervised machine learning methods for analyzing spatially distributed seismic damage after an earthquake at the regional scale. Additionally, we aim to quantify uncertainties in building damage assessment at the building-level, using in-situ earthquake data recorded at the top and the bottom floors of buildings.