Seismic Ground Motion - Dynamic Behavior of Concrete and Seismic Engineering - page 120

modulus decreases and the damping increases when the imposed strain increases:
G = G max 1./(1+J/J r ). Such behavior might be expected to lead to a decrease of both
frequencies and resonance associated amplifications, as well as reductions in the
high-frequency content (maximum acceleration, especially).
The question is knowing where the strains imposed by an earthquake are located
(they can quite often exceed 10 -3 , and sometimes reach 10 -2 ) with regard to the
critical strain values (J r ). Seismologists and geotechnicians disagree on this point,
the former believing that accelerometer observations were best explained by
reference to linear visco-elastic behaviors, whilst the latter measured (J r ) values
under 10 -3 under laboratory conditions. In the last two decades, observations have
tended to reconcile both points of view, owing to the simulation of far less non-
linear behaviors under lab conditions, notably for very plastic grounds [VUC 91],
and the observation of non-linear in situ effects, especially in sand layers. Some
quantitative disagreements still exist, since even the best accelerometer data seems
to reveal a slightly to appreciably less non-linear behavior than that predicted by
numerical models derived from laboratory measurements [BON 03].
It is important to grasp the reason for inconsistencies between in situ and
laboratory observations, particularly for seismic zones such as France. Currently,
French regulation (PS92) authorizes a 20% decrease of the high frequency content in
soft ground, implicitly assuming highly non-linear grounds, and has proved to be
rather conservative. The EC8 recommendations allow high frequency amplifications
of up to 80%, whilst the latest American propositions reach 200%.
3.5. Conclusions
Strong motion seismology is a relatively new subject which has evolved a great
deal over the last decades, advances often being due to the questions raised by
“abnormal” damage and intensity observations made during destructive earthquakes,
generally via accelerometer recordings. Whilst estimating seismic motions was
mainly empirical in the last century, many models that were satisfactory from a
physical point of view emerged in the following decades, and have been
progressively adjusted and calibrated to real time instrument measurements.
The percolation time between “normal observations” and their use in seismic
engineering practice remains quite long (about a decade and even more for
conventional regulations) for the following reasons: if we simplify in the extreme,
the seismic force F which a building will have to withstand is proportional to three
terms: F v a z . S' a (T, ] )/Qs , where: