Earthquake risk management of civil infrastructure: integrating soft and hard risks - Seismic Risk Analysis and Management of Civil Infrastructure Systems - page 236

issues addressed by practitioners. We have to include the realities of human
and organisational issues and relationships with wider society, which are
utterly important in engineering practice and dominate the causes of engi-
neering failures, accidents and disasters (Blockley, 1980, 2010; Reason, 1990;
Turner and Pidgeon, 1998). Grundy (2011) has proposed a formal introduc-
tion of a disaster limit state which goes beyond an ultimate limit state to
consider risk reduction strategies that are not normally considered to be
engineering but a wider part of disaster management.
We will use the term hard system for the inanimate physical things, such
as beams, columns, bridges, dams and reservoirs. We will use the term soft
system where people and social systems are involved. Hard systems are
physical systems that are commonly said to be 'objective' in that they are
supposed to be independent of the observer and hence the same for all of
us. Reliability analyses are almost always technical and only include uncer-
tainties that can be modelled as random or stochastic variations in param-
eters of demand and capacity or load and strength of hard systems - we will
call that hard system parameter uncertainty . As such, they are sophisticated
and more coherent versions of much simpler safety factors.
Analyses of behaviour and performance are based on the physical science
of engineering including models that are tested in laboratory under pre-
cisely defi ned conditions (e.g. knife edge supports) with precisely known
parameters (load and material properties). They are also tested in practice
through the successful performance of the millions of applications that give
us confi dence that they are dependable. Unfortunately the conditions under
which these structures in practice operate are rarely precisely known and
they are rarely required to perform at levels close to their limit states. Nev-
ertheless, as long as we are aware of the contexts in which these models
work, they are very dependable and the basis for many successful structures.
Of course failures (which are fortunately rare) are important learning
points, since we know then that our methods have been demonstrably
wrong or falsifi ed. Any differences between theoretical and actual behav-
iour, even when we know the parameters quite precisely, contribute to hard
systems model uncertainty . These kinds of uncertainties relate to any differ-
ences in behaviour between theoretical predictions and actual behaviour.
Laboratory tests reveal some (usually small) differences even though test
conditions and parameters are precisely controlled. However, the major
uncertainties arise due to differences in modelling assumptions and the
structure as-built. For example, we may assume simple supports but in
practice the supports may be designed and built such that they carry some
bending moments (for example, welded end plates on steel beams). Unfor-
tunately measurements of actual 'as-built' structural behaviour in operation
are rare, although developments in better and cheaper instrumentation
systems are leading to more tests and studies on full-scale structures. The