Experimental Methods in Earthquake Engineering - Dynamic Behavior of Concrete and Seismic Engineering - page 180

' and ' are constant. Then the dilatation factor can be
modified interactively just by modifying s n during the experiment. Actually, we
always try to decrease O as much as possible while keeping the control error within a
satisfactory interval (typically < 50 P m), which indeed depends on the nature and the
complexity of the test. In any case, as with the conventional PSD method, it is
important to keep the errors as low as possible, and the motion regular. The absence
of relaxation imposed on the structure by the PSD method definitely has a big part to
play in it. Let us also note that as the reaction force is now taken into account with a
high frequency (500 or 1,000 Hz), an excellent noise filter of this analog
measurement is thus introduced, which is far better than in the case of the
conventional PSD method where the number of measuring points of the force taken
into account is much lower.

For a given test,

st

In practice, the values of O can be close to the unit for a simple test (one dof in
the lab with potential sub-structuration) with negligible forces of inertia in the
experimental device. They can be much higher, up to 400 for a complex tested
structure.

5.3.2. Implementation at ELSA

When tests with several jacks have to be carried out, it is necessary to achieve
perfect synchronization of each jack. This requires some refinement in the
architecture of the control system. The difficulty involves coordinating and carrying
out for a specific time period of the control system (1 or 2 ms) all the tasks
associated with a cycle of the PSD method: measuring, calculation of the movement
and control of the displacement. Therefore, new control units, more powerful and
more modular have been introduced, communicating with the computer in charge of
the integration of the equations of motion, owing to a new broadband system.

As shown in Figure 4.5, each control unit consists of 3 modules integrated on the
same card: signal processing and storing, acquisition and control, communication
and exchange of data. These cards can then be plugged on the same single
motherboard and work together owing to the data bus of the host computer, as well
as with its processor that is in charge of the integration algorithm. All these
processors (the slaves checking while the master is integrating) operate in real time
with an appropriate operating system (TNT on 32 bits), synchronized by a clock.
Moreover, as a major task via the network, the master processor also provides the
interface with other computers working with more standard operative systems
(Windows NT, 2000 or XP).

There lies another key point of the new architecture: easy data access. In using
the possibilities of active object distribution via the network (DCOM technology by