Gaussian process regression for seismic fragility assessment of building portfolios
Section snippets
Introduction and motivations
Seismic fragility is quantitatively expressed as the conditional probability that a structure will reach or exceed a specified level of damage (or damage state, DS) for a given value of a considered earthquake-induced ground-motion intensity measure (IM). Fragility relationships describe such a conditional probability for increasing values of the ground-motion IM, taking the form of cumulative distribution functions (CDFs). Typically, a lognormal model characterised by two parameters - median
Commonly adopted metamodeling techniques
This section briefly reviews some of the commonly-adopted metamodeling techniques, also compared in Table 1. A thorough examination of all the available approaches in the literature is outside of scope of this paper. The desired features of the metamodel (in relation to fragility modelling) are described, highlighting the ones leading to the choice of the GP regressions.
The Response Surface Method (RSM; [17]) models the relationship between several explanatory (input) variables and one or more
Methodology
This study aims 1) to show the feasibility of using GP regressions to explicitly consider building-to-building variability in seismic fragility assessment of building portfolios; 2) to provide practical guidance on how to develop (i.e. train and validate) GP regressions for case-study applications. The selected parameters to be surrogated are the median and the dispersion (i.e. logarithmic standard deviation) of the fragility functions defined for four different damage states (Fig. 1). The
Illustrative application
The case-study buildings selected here represent seismically-deficient RC school buildings, defined based on a large data-collection exercises performed by the authors [51], involving rapid visual surveys for over 200 school buildings. Specifically, a rapid visual survey was carried out for each building to collect administrative (i.e. location, year of construction/retrofit, number of students, teachers), geometric (i.e. member dimensions) and mechanical data (i.e. structural details, nominal
Estimated database of fragility functions
Fragility analysis is carried out for each building realisation, based on both DoE and MCS sampling techniques. Fig. 5 shows all the results based on time-history analysis. The period range for the definition of is defined as , where is the minimum first-mode period for the entire database while is the maximum. This allows one to obtain consistent results (i.e. using the same ground-motion IM for all the building realisation), with a minimum
Conclusions
This paper proposed using Gaussian process (GP) regressions to develop flexible and accurate metamodels that can be used to account for building-to-building variability in simulation-based seismic risk assessment of building portfolios. This involves analysing a small number of samples (i.e. building realisations of the same class) and subsequently fitting GP regressions to the seismic fragility parameters obtained through cloud-based non-linear time-history analysis of each building
Acknowledgements
This study was performed in the framework of the “INSPIRE: INdonesia School Programme to Increase REsilience” and “i-RESIST: Increasing REsilience of Schools in Indonesia to earthquake Shaking and Tsunami” projects, funded by the British Council through the Newton Institutional Links scheme and Research England through the UCL Global Challenges Research Fund (GCRF) Small Research Grants scheme. Additional funding was received from the European Union’s Horizon 2020 research and innovation
References (59)
- et al.
Incorporating modeling uncertainties in the assessment of seismic collapse risk of buildings
Struct Saf
(2009) - et al.
Seismic risk of R.C. building classes
Eng Struct
(2007) - et al.
Analytical seismic assessment of RC dual wall/frame systems using SLaMA: Proposal and validation
Eng Struct
(2019) - et al.
Efficient input-output model representations
Comput Phys Commun
(1999) - et al.
Stochastic response of reinforced concrete buildings using high dimensional model representation
Eng Struct
(2019) - et al.
Artificial neural network based multi-dimensional fragility development of skewed concrete bridge classes
Eng Struct
(2018) - et al.
Surrogate modeling and failure surface visualization for efficient seismic vulnerability assessment of highway bridges
Probabilistic Eng Mech
(2013) - et al.
BEA: An efficient Bayesian emulation-based approach for probabilistic seismic response
Struct Saf
(2018) - et al.
FRACAS: A capacity spectrum approach for seismic fragility assessment including record-to-record variability
Eng Struct
(2016) - et al.
Accounting for spectral shape in simplified fragility analysis of case-study reinforced concrete frames
Soil Dyn Earthq Eng
(2019)