Environmental Engineering Reference
In-Depth Information
which suggested that a 10% chance of failure can
be expected at the design load (reflecting the var-
ious safety factors inherent within traditional de-
sign) and a 50% chance of geotechnical failure
when the factor of safety is equal to 1. These
figures are useful rules of thumb that can be used
to give confidence in the results of the more com-
plex full reliability analysis.
be extended and made bespoke to specific assets
through user-defined LSEs (either based upon em-
pirical formulae not yet coded within the Failure
Mode Library or based on the emulation of more
complex models).
.
A database of parameters and variables - for a
given flood defence structure, values must be sup-
plied for each parameter and variable required by
the relevant LSEs. A value may be fixed or spec-
ified as a statistical distribution with associated
parameters.
.
A Monte Carlo simulation - a large sample of
input variables (strength and load) are generated
and the annual probability of failure, conditional
failure probability (where the hydraulic loading
conditions are specified as fixed variables and then
the strength variables systematically varied) and
other related statistics calculated (van Gelder
et al. 2008). The number of simulations required
to achieve a converged estimate of the probability
of failure, and thus the calculation time, depend on
the chance of asset failure. Most structures in
coastal and river engineering, for example, exhibit
a relatively high probability of failure (i.e. a rela-
tively low reliability - typically an annual proba-
bility of failure
>
0.005) compared to structures in
other industries where reliability analysis is rou-
tinely applied (e.g. for the structural components
of a nuclear power plant or mechanical compo-
nents of an aeroplane the typical reliability will be
much higher,
<
0.0001). This presents Monte
Carlo simulation as a viable and flexible numer-
ical integration tool in the context of the majority
of flood defence assets. In the case of complex
failure surfaces, where the response of the struc-
ture exhibits discontinuities, run times can in-
crease to ensure such discontinuities are
captured. In such cases, innovative sampling tech-
niques (e.g. importance sampling) are techniques
that could be usefully employedwithinRELIABLE
to minimize run times (van Gelder et al. 2008).
In generating the asset-specific fragility infor-
mation it is important to understand how these
methods relate to more traditional assessment
methods (based on partial factors of safety). An
interesting comparison was completed as part of
the Thames Estuary studies (Simm et al. 2008),
System Analysis and Attributing Risk to
Individual Assets
Risk-based management requires a comprehen-
sive consideration of the sources, pathways and
receptor impacts. In the context of asset manage-
ment, this implies that the asset manager must
look beyond the performance of individual
assets to understand the behaviour of the asset
system, and the variation in loading and conse-
quences. System risk analysis, based on the
Source-Pathway-Receptor concept (DETR 2000;
Sayers et al. 2002) and methods that sample mul-
tiple asset failure combinations (system states)
and loading conditions together with the associ-
ated impact, are now well established (Hall
et al. 2003; Sayers and Meadowcroft 2005;
Gouldby et al. 2008, 2009). Within England and
Wales the so-called 'RASP High Level Method'
(HR Wallingford 2009) is used in the National
Flood Risk Analysis (NaFRA; Steel et al. 2009) and
is currently being implementedwithin theModel-
ling Decision Support Framework (MDSF2;
Surendran et al. 2008; Environment Agency 2009)
and the Performance-based Asset Management
System (PAMS; Simm et al. 2006; Environment
Agency 2010) being developed by the Environment
Agency.More detailed systemanalysis techniques
that relax some of the assumptions within the
RASPHigh Level Method are also being developed
(e.g. within the Flood Risk Assessment under
Climate change, FRACAS; Gouldby et al. 2009).
Flood risk system models are typically used to
provide a means of quantifying the spatial distri-
bution of risk within the floodplain. For asset
management purposes, however, it is desirable to
identify those defence sections that make a sig-
nificant contribution to the risk. This aids the
