Experimental Methods in Earthquake Engineering - Dynamic Behavior of Concrete and Seismic Engineering - page 196

For random noise loading, we start from a flat spectrum in g²/Hz. The software
generates a random signal whose frequency band energy corresponds to that of the
spectrum defined for the tests. During the test, driving software checks that the
energy level remains constant within each frequency range throughout. Acquisition
is carried out using a certain number of samples (a minimum of 64 to obtain a 5%
accuracy). The duration of the samples depends on the analysis frequency band: it is
longer when we want to analyze low frequencies. For each sample, a PSD is
calculated and an average is obtained for all samples. Comparing the PSDs of the
sensors with the PSD of the table sensor enables the identification of the resonance
frequencies. This kind of test is generally less disadvantageous for reinforced
concrete structures: unlike a sinus test, it stays at the resonance frequency for a
shorter time and the maximum amplitudes reached are less important.
The time history tests involve applying a given seismic load to the table and in
reproducing it as well as possible. The load can correspond to any of the following:
- a real earthquake recorded on site (real accelerogram);
- acceleration derived from pre-dimensioning calculation;
-acceleration generated from the spectrum of oscillators (synthetic
accelerogram) for a given damping.
5.9. Conclusion
In conclusion, the advantages of shaking table tests are as follows:
- they allow simulation of inertia loads on a structure with a distributed mass;
- they allow tests to be carried out with vertical loads;
- they reveal some dynamic behaviors (vertical normal loads due to the opening
and closing up of cracks);
- it is not necessary to use calculations to take the structure damping into
account since they give the real response of the structure.
However, their drawbacks include the following:
- limitation of the loaded mass, which dictates carrying out tests on reduced
scale models, which may not be truly representative;
- the stiffness of the table influences the eigenfrequencies of the model;
- the soil-structure interaction is difficult to take into account (generally, the
table is considered as hard soil (rock-type));