II. How Could System Identification Be Applied to Earthquake Engineering?

System identification serves as a key component in structural control, structural health monitoring and
damage detection. For earthquake engineering in general, system identification, which is based on
real-world dynamic data measurements, facilitates structural modeling for earthquake simulation,
analysis and design. Apart from this, many analytic and numerical tools in system identification that are
powerful in data processing and data interpretation are invaluable to the earthquake engineering
community.

Benefits from applying the results of system identification and/or “borrowing” the approaches and
techniques in system identification to earthquake engineering may be summarized in the following aspects:

1. Modeling structures based on real-world measurements rather than design assumptions;

2. Introducing nonlinearities into structural modeling;

3. Applying probabilistic methods for performance-based design and seismic risk assessment;

4. Analyzing and simulating earthquake responses;

5. Monitoring structural health for seismic protection, and   

6. Providing data processing and interpretation tools.

1. In system identification, models are derived from real-world dynamic measurements.
Thus the models are supposed to reflect the real status of structures rather than being derived from
assumed and/or idealized values of the key parameters. Although tremendous challenges exist in
correctly modeling structures based on input-output data sets, for existing structures especially those
aging infrastructures, being able to capture the current properties of the structures is critical in predicting
their earthquake performance and survivability.

2. Nonlinearities are considered in some system identification techniques.
Under a strong ground shaking caused by an earthquake, it is possible for a structure to behave
in an inelastic and nonlinear range. Powerful tools in handling linear vibrations, such as Fourier transform
and modal analysis are normally not applicable to nonlinear problems. The need of evaluating and
considering inelastic seismic analysis procedures has been pointed out by the ACT-55 report
(which can be downloaded from the Applied Technology Council website). System identification
offers tools to model and analyze nonlinearities for the applications in earthquake engineering. See details.

3. Probabilistic methods are applied widely in system identification techniques.
Probabilistic methods play an important role in earthquake engineering due to the random nature of
earthquake excitations and the uncertainties involved in structural systems due to construction errors,
material aging, and environmental changes. Similarly, probabilistic methods are widely studied in
system identification to deal with the complexities of the real-world uncertainties in systems and loads.
The applications of probabilistic methods in earthquake engineering include running
Monte Carlo simulation,
generating artificial ground motions, producing fragility curves, and performing earthquake risk assessment.
See details.

4. Models can be used to simulate global earthquake response of a structure for a given/specified
earthquake ground motion. Based on a limited number of sensors where the data measurements
are collected, system identification provides mathematical models of structures. Earthquake simulation
thus becomes a forward problem utilizing a specified ground motion time history and the model of
the structure obtained from system identification.

5. Paired with sensor technology, system identification can provide an efficient means in monitoring
structural health for seismic protection. See details.

6. Various numerical schemes and signal processing methods developed or applied in system identification
are highly relevant to earthquake engineering applications. Earthquake excitations and responses can be
simulated numerically or measured from a laboratory setting or from site. It can become a routine to handle
long data sets in earthquake engineering. System identification provides various powerful tools in
data processing and data interpretation. Apart from the widely applied modal analysis based approaches,
there are various techniques handling nonlinear systems and non-stationary data measurements associated
with strong earthquake motions. Among them, there are artificial neural networks, wavelets, and
Hilbert Huang transform
. The last one is an emerging technique, and the first two are considered as
black-box methods. Clicks the links for details.