Artificial neural networks approach to predict principal ground motion parameters for quick post-earthquake damage assessment of bridges
Having a quick but reliable insight into the likelihood of damage to bridges immediately after an earthquake is an important concern especially in the earthquake prone countries such as New Zealand for ensuring emergency transportation network operations. A set of primary indicators necessary to perform damage likelihood assessment are ground motion parameters such as peak ground acceleration (PGA) at each bridge site. Organizations, such as GNS in New Zealand, record these parameters using distributed arrays of sensors. The challenge is that those sensors are not installed at, or close to, bridge sites and so bridge site specific data are not readily available. This study proposes a method to predict ground motion parameters for each bridge site based on remote seismic array recordings. Because of the existing abundant source of data related to two recent strong earthquakes that occurred in 2010 and 2011 and their aftershocks, the city of Christchurch is considered to develop and examine the method. Artificial neural networks have been considered for this research. Accelerations recorded by the GeoNet seismic array were considered to develop a functional relationship enabling the prediction of PGAs. http://www.nzsee.org.nz/db/2013/Posters.htm