Machine Learning for City-Scale Building Information Modeling

The speed and accuracy of seismic loss estimation are central to effective post-earthquake emergency response. Inadequate emergency response can increase the number of casualties by a maximum factor of 10, which suggests the need for research on rapid earthquake shaking damage and loss estimation. For maximum utility, these estimates need to be conducted at as fine a scale as is practically possible. To this end, we propose a framework that leverages the advantages of recent breakthroughs in remote sensing, non-linear structural dynamics and Artificial Intelligence (AI) for shaking damage to buildings at a large (regional) scale. The framework provides a method to quickly build a building inventory of a given region, which enables rapid building damage estimation based on nonlinear dynamic analyses