摘要(英) |
Located at the
active arc-continent collision region between the Luzon arc of the
Philippine Sea plate and the Eurasian plate, Taiwan is subject to
frequent earthquakes. Because of their devastating potential, there is a
great interest in predicting the location and time of large earthquakes.
Although a great deal is known about where earthquakes are likely to
occur, there is currently no reliable way to predict the time when an
event will occur in any specific location. However, the damages caused
by them can be greatly reduced with proper structural design using safer
seismic code. In this regard, dynamic behavior of structures under
earthquakes should be considered in the process of design. In order to
realize the dynamic behavior of structural systems, we can determine the
dynamic models and parameters by system identification techniques. In
the past few decades, many optimization techniques have been employed
for system identification problems. Most of the identification methods
mentioned above are calculus-based search method. They are performed by
point-to-point search strategy. A good initial guess of the parameter
and gradient or higher-order derivatives of the objective function are
normally required. There is a possibility to fall into a local minimum
rather than the global minimum. On the contrary, genetic algorithms
(GAs) are optimization procedures inspired by natural evolution. They
model natural processes, such as selection, recombination, and mutation,
and work on populations of individuals instead of single solutions. In
this way, the algorithms are parallel and global search techniques that
search multiple points, so they are more likely to obtain a global
solution. While the GA method has been developed as a powerful
search tool in a global solution space, it is not necessarily efficient
in fine-tuning for local convergence particularly when the search domain
is large. In order to accelerate the convergence to the optima
solutions, a hybrid identification strategy, combining GA and local
search technique such as Gauss-Newton method is proposed in this study.
The proposed algorithm is explored by comparing the results of the
predicted response with the measured response for both the SDOF
linear/nonlinear system and the MDOF linear/nonlinear system with or
without noise contamination. Finally, the hybrid computational strategy
is also applied to the Taiwan Electricity Main Building using records
from the 331 earthquake (2002). The comparison is made between the
predicted acceleration and the measured one for each case. |