Civil Engineering Reference
In-Depth Information
Seismic activity of the seismogenic sources/faults (modelled through
magnitude-recurrence laws, probability distributions for epicentre loca-
tions or rupture area, etc.).
Local seismic intensities at the sites (modelled through ground-motion
prediction equations, spatial correlation models, cross-IM correlation
models, and site amplifi cation models).
Physical damageability of the components of the infrastructure
(modelled by fragility models for structural and non-structural
components).
Functional consequences of the physical damage.
Socio-economic consequences of physical damage (modelled by proba-
bilistic cost models).
Further uncertainty enters the model because alternative models are avail-
able for each of the above items, and the corresponding parameters are
themselves characterised by statistical uncertainty. Hence, for instance, in
the seismic activity model the sources' boundaries, or the parameters of the
associated recurrence law, like lower and upper magnitudes, can be defi ned
with some variability. Denoting with
x
, the vector collecting all random
variables employed to model the uncertainties in the problem, probabilistic
assessment of the model requires the joint probability density function
f
(
x
).
The way this joint density is employed to assess performance metrics such
as those described in Section 18.6.3 depends upon the adopted reliability
method.
18.7.2 Available methods
Available reliability methods are either simulation-based or non-simula-
tion-based. Both have been applied to the analysis of infrastructural systems.
The model is designed not to take a specifi c position on this matter; thus
an
Analysis
class is composed of both
Simulation
and
nonSimulation
abstract classes. Pragmatically, the implemented portion of the model
includes only simulation-based methods, and due to the reasons to follow,
no non-simulation-based methods are implemented. In particular, some of
the promising methods/approaches include the so-called matrix system reli-
ability method (MSRM) (Song and Kang, 2009; Song and Ok, 2010) and
the model of bayesian networks (BN) (Jensen and Nielsen, 2007; Straub
et
al.
, 2008; Der Kiureghian, 2009; Straub and Der Kiureghian, 2009; Bensi
et
al.
, 2009).
The fi rst method appears to reach a limit when dealing with capacitive
(fl ow) modelling of networks. Applications, to the knowledge of the authors,
are limited to connectivity problems, as in Kang
et
al.
(2008). The second
method seems very promising in perspective, especially for dealing with the
Search WWH ::
Custom Search