In this study, neural network models improved by genetic algorithm were employed to estimate peak ground acceleration (PGA) at seven metropolitan areas in the island of Taiwan, which is frequently subject to earthquakes. By considering a series of historical seismic records, and using the seismic design value in the current building code as the evaluation criteria, two metropolitan areas, Taichung and Chiayi, were identified by computational results as having higher estimated horizontal PGAs than the recommended design values. The approach implemented in this study provides a new and good basis for solving this type of seismic problems in the region studied.
Actual seismic records usually involve a very high nonlinear data set, which may require a tedious work to access by conventional statistical and vibration analysis. Alternatively, the present study develops an improved artificial neural network (ANN) model for evaluating the current seismic zone divisions in Taiwan's standard using the global search capability of genetic algorithm (GA). This model (GA+ANN) predicts the key factor of peak ground acceleration (PGA) using as inputs the recorded values of actual earthquake parameters including magnitude, epicenter distance, and focal depth. Results are presented to show that this model exhibits an improved generalization capability, with acceptably high coefficient of correlation and low root mean square error between estimations and records. In addition, four locations out of twenty-four subdivision zones in total are identified by the model as having a potential for experiencing a higher horizontal PGA than that of the design value in the standard. Furthermore, the equation PGA = 0.44 lexp(-0.020£>/), developed by curve fitting, is presented for approximately describing the relationship between horizontal PGA and focal distance (Dj). The method employed provides a new approach to treat this type of nonlinear problem, and the information obtained provides a good basis for further research in this region.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.