Page 183 - Dynamic Loading and Design of Structures

Page 158

The selected record is then processed iteratively by multiplying the Fourier amplitudes (see
eqn 4.11) by the corresponding average of the ratios of target Spv values to the Spv values
calculated for the initially selected record, with a view to better matching the target spectrum.
In the ‘seismological method’, simulated accelerograms are generated by modelling the
source and travel path mechanisms. The method generally involves two steps (Hu et al.,
1996):

●define the ground motion at the site due to an ‘element’ of earthquake source or fault
rupture; planar sources are divided into a number of elements
●sum up the contributions of motions due to all elements, in the time domain.

A detailed discussion of this method, which is less common than the previous one, falls
outside the scope of this book. References to the pertinent literature can be found, for instance,
in Clough and Penzien (1993) and Hu et al. (1996).

Code treatment
All the aforementioned types of accelerograms are generally allowed as input for time history
analysis in EC8, which, however, appears to promote spectrum compatible records, generated
using the elastic response spectrum as the target. The duration of the records must be
consistent with the characteristics (M, R, etc.) of the earthquake underlying the establishment
of the design α A minimum of five records is required for time history analysis, which
g.
should be enough to provide a stable statistical measure of the response; additional rules are
given in EC8 regarding the allowable difference between the mean spectrum of these records
and the code spectrum.
Whereas spectrum compatible records are an attractive choice of dynamic input, in the
sense that scaling is not required and code requirements are imposed in a rather
straightforward way, care is required in their construction to avoid over conservatism as well
as inconsistencies. Referring to the previously described EC8 procedure, it is emphasized that
it is the mean of the response spectra of artificial motions that should match the design
spectrum, rather than each individual spectrum. In practice what is commonly done is that the
elastic design spectrum is used as the target for all records (i.e. each spectrum matches closely
the design one); this is inconsistent with the very nature of the design spectrum which does
not represent a particular ground motion but rather envelopes the spectra of several motions
generated from different sources and at different distances from the site. As shown by Naeim
and Lew (1995), design spectrum compatible motions may represent velocities, displacements,
and energy content which are very unrealistic; as a result their use in inelastic time history
analysis may lead to unreliable estimates of design displacement demands.
Eurocode 8 also allows the use of recorded (natural) accelerograms, as well as of
accelerograms generated by simulation of the source and travel path effects (seismological
method). A minimum of three records is required, to be scaled to the design