Cardiff University | Prifysgol Caerdydd ORCA
Online Research @ Cardiff 
WelshClear Cookie - decide language by browser settings

A consensus-based approach for structural resilience to earthquakes using machine learning techniques

Cere, Giulia 2019. A consensus-based approach for structural resilience to earthquakes using machine learning techniques. PhD Thesis, Cardiff University.
Item availability restricted.

[img]
Preview
PDF - Accepted Post-Print Version
Download (6MB) | Preview
[img] PDF (Cardiff University Electronic Publication Form) - Supplemental Material
Restricted to Repository staff only

Download (131kB)

Abstract

Seismic hazards represent a constant threat for both the built environment but mainly for human lives. Past approaches to seismic engineering considered the building deformability as limited to the elastic behaviour. Following to the introduction of performance-based approaches a whole new methodology for seismic design and assessment was proposed, relying on the ability of a building to extend its deformability in the plastic domain. This links to the ability of the building to undergo large deformations but still withstand it and therefore safeguard human lives. This allowed to distinguish between transient and permanent deformations when undergoing dynamic (e.g., seismic) stresses. In parallel, a whole new discipline is flourishing, which sees traditional structural analysis methods coupled to Artificial Intelligence (AI) strategies. In parallel, the emerging discipline of resilience has been widely implemented in the domain of disaster management and also in structural engineering. However, grounding on an extensive literature review, current approaches to disaster management at the building and district level exhibit a significant fragmentation in terms of strategies of objectives, highlighting the urge for a more holistic conceptualization. The proposed methodology therefore aims at addressing both the building and district levels, by the adoption of scale-specific methodologies suitable for the scale of analysis. At the building level, an analytical three-stage methodology is proposed to enhance traditional investigation and structural optimization strategies by the utilization of object-oriented programming, evolutionary computing and deep learning techniques. This is validated throughout the application of the proposed methodology on a real building in Old Beichuan, which underwent seismically-triggered damages as a result of the 2008 Wenchuan Earthquake. At the district scale, a so-called qualitative methodology is proposed to attain a resilience evaluation in face of geo-environmental hazards and specifically targeting the built environment. A Delphi expert consultation is adopted and a framework is presented. To combine the two scales, a high-level strategy is ultimately proposed in order to interlace the building and district-scale simulations to make them organically interlinked. To this respect, a multi-dimensional mapping of the area of Old-Beichuan is presented to aid the identification of some key indicators of the district-level framework. The research has been conducted in the context of the REACH project, `vi investigating the built environment’s resilience in face of seismically-triggered geo-environmental hazards in the context of the 2008 Wenchuan Earthquake in China. Results show that an optimized performance-based approach would significantly enhance traditional analysis and investigation strategies, providing an approximate damage reduction of 25% with a cost increase of 20%. In addition, the utilization of deep learning techniques to replace traditional simulation engine proved to attain a result precision up to 98%, making it reliable to conduct investigation campaign in relation to specific building features when traditional methods fail due to the impossibility of either accessing the building or tracing pertinent documentation. It is therefore demonstrated how sometimes challenging regulatory frameworks is a necessary step to enhance the resilience of buildings in face of seismic hazards.

Item Type: Thesis (PhD)
Date Type: Completion
Status: Unpublished
Schools: Engineering
Uncontrolled Keywords: Building Resilience; Seismic hazard; Deep learning; Performance-based analysis; Optimization; Reinforced Concrete frame.
Date of First Compliant Deposit: 13 March 2020
Last Modified: 13 Mar 2020 10:17
URI: http://orca.cf.ac.uk/id/eprint/130384

Actions (repository staff only)

Edit Item Edit Item

Downloads

Downloads per month over past year

View more statistics