A computational framework for modelling post-earthquake building damage, loss of functionality and recovery

Large earthquakes often cause significant socio-economic disruptions to communities that last beyond physical damage to buildings and infrastructure. In both New Zealand and international contexts, this has prompted a shift in seismic design paradigm from a sole focus on life safety to approaches that also limit damage and support timely recovery. However, accurate prediction of how damaged buildings regain functionality, and at what pace, remains a complex challenge requiring large empirical data and advanced modelling techniques. The 2010/11 Canterbury Earthquake Sequence (CES) in New Zealand, with its comprehensive empirical dataset, provides an opportunity to address this challenge. By applying machine learning (ML) techniques and probabilistic methods, this research aims to develop a computational framework for modelling building damage, loss of functionality and recovery post-earthquake. The study first undertakes a systematic review to trace the evolution of seismic design priorities towards functionality-centred objectives. Multiple ML algorithms are then evaluated for rapid damage assessment (RDA), enabling identification of a contextually robust model and minimal predictor sets consistent with New Zealand inspection protocols. The Random Forest (RF) model, selected for its strong RDA performance, is subsequently adapted for functionality assessment to identify critical components, map damage propagation pathways, and generate probabilistic functionality estimates at scale. To enhance downtime estimation, a hybrid framework integrates Bayesian inference with ML to refine recovery parameters and forecast community-level recovery trajectories. Human resource constraints are further examined using a Dynamic Stochastic Queuing (DSQ) model to quantify the influence of mobilisation patterns and repair sequencing on restoration timelines. The resulting framework integrates engineering, socio-economic, and operational dimensions of post-earthquake recovery, linking micro-scale building performance data with macro-scale recovery processes. By combining rapid damage assessment, functionality evaluation, recovery forecasting, and workforce constraint modelling, it enables reproducible analysis of how design provisions, inspection protocols, and resource mobilisation shape recovery pace. This work provides evidence-based research to inform the development of New Zealand’s Low-Damage Design (LDD) guidance and functional recovery initiatives. It also offers a generic approach to ML modelling for resilience-oriented design and recovery planning.

Name thesis:degree_name

PhD

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Civil Engineering

Grantor dc:publisher

ResearchSpace@Auckland

Year dc:date.issued

2025

Author dc:creator

• Li, Lianyan

Advisors dc:contributor.advisor

• Chang-Richards, Alice
• Boston, Megan
• Elwood, Ken

• Post-earthquake building recovery
• machine learning
• building functionality
• downtime modelling
• seismic resilience
• functional recovery
• 2010-2011 Canterbury Earthquake Sequence
• New Zealand

Statement dc:rights

• Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Licence dc:rights.uri

• https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm

Handle dc:identifier.uri

https://hdl.handle.net/2292/74707

OAI identifier oai:identifier

oai:researchspace.auckland.ac.nz:2292/74707