LIQUEFACTION POTENTIAL ASSESSMENT USING MULTILAYER ARTIFICIAL NEURAL NETWORK

Abstract

In this study, a low-cost, rapid and qualitative evaluation procedure is presented using dynamic pattern recognition analysis to assess liquefaction potential which is useful in the planning, zoning, general hazard assessment, and delineation of areas, Dynamic pattern recognition using neural networks is generally considered to be an effective tool for assessing of hazard potential on the basis of established criteria. In this paper, the classification operation, in which an input pattern is passed to the network and the network produces a representative class as output, is considered for evaluation of liquefaction hazard potential. The application of Multilayer Artificial Neural Network for the prediction of liquefaction was examined in the northwest of Iran (Gilan Plain). The study area suffered acatastrophic earthquake in June 1990 and most of the damage to lifeline facilities and structures in urban areas was brought about by liquefaction phenomena. The simulated results by multilayer artificial neural network in this study revealed the high capability of this method to predict the liquefaction potential of soils