Quantifying earthquake-induced damage, structural response, and resilience across large portfolios of buildings poses a significant challenge, primarily due to the extensive computational resources required. In postdisaster scenarios, rapid decision-making is crucial for enabling the swift recovery of affected urban areas. This study introduces a novel urban cluster earthquake resilience (UCER) framework, which leverages advanced machine learning techniques, including t-distributed stochastic neighbor embedding (t-SNE), hierarchical density-based spatial clustering of applications with noise (HDBSCAN), and K-nearest neighbors (KNN), to assess the resilience of building clusters efficiently. These clusters are represented by surrogate models that encapsulate the key characteristics of a diverse portfolio of 23,420 residential RC and masonry structures in an idealized urban setting. The primary objective is to create a robust, data-driven method that enables the rapid, a priori determination of building resilience in disaster-affected environments. This framework empowers decision-makers to improve construction strategies and emergency planning without relying on computationally expensive analyses. The study demonstrates that by using the UCER framework, resilience indices can be efficiently generated for RC and masonry buildings across multiple disaster scenarios to offer valuable insights without the need for extensive calculations.
Chavez et al. (2026) studied this question.