Liquefaction potential assessment in layered soils by probabilistic neural networks based on the Kocaeli, Turkey earthquake

Ural D., Bayrak M., Saygili G.

International Conference on Cyclic Behaviour of Soils and Liquefaction Phenomena, Bochum, Germany, 31 March - 02 April 2004, pp.505-514 identifier

  • Publication Type: Conference Paper / Full Text
  • Doi Number: 10.1201/9781439833452.ch61
  • City: Bochum
  • Country: Germany
  • Page Numbers: pp.505-514
  • Istanbul Technical University Affiliated: No


Emerging artificial intelligence tools, namely neural networks, are reliable and cost efficient alternatives that have been successfully applied to several geotechnical engineering problems. This paper will present two artificial intelligence models using a specific type of neural network, the probabilistic neural network (PNN) applied to a specific soil site. The liquefaction phenomenon has been modeled effectively utilizing back propagation and general regression neural networks, GRNN. The probabilistic neural network methodology is utilized to assess the liquefaction potential in layered soils by using actual field records obtained from the province of Kocaeli, situated in the northwestern part of Turkey. A magnitude 7.6 earthquake occurred in Kocaeli at 3:02 AM on August 17, 1999 and soon after the earthquake, soil collection and in-situ testing was performed in regions that had liquefied.