P. Alvanitopoulos, I. Andreadis, and A. Elenas (Greece)
Pattern classification, neural networks, seismic parameters, seismic damages.
It is well-known that the natural phenomenon of an earthquake is unpredictable both in time and intensity. Consequently, the damages on human constructions produced by an earthquake are also unpredictable. The aim of this paper is to classify the structural and architectural damages suffered after an earthquake. The interest is obvious for several reasons such as public safety, economical recourses management, infrastructure and urban planning. In this paper a new classification method based on artificial neural networks is proposed. A set of 200 accelerograms were used. Two artificial neural networks, one for each type of damages, have been proposed. The neural network is a Back Propagation one with a supervised learning scheme for the training process. Correct classification rates up to 94.5% and 93.5% for the structural and architectural damages, respectively, are achieved.
Important Links:
Go Back
