AFSs-RBF neural network for predicting earthquake-induced liquefaction of light loam

In this study, AFSs-RBF neural network, based on the adaptive fuzzy systems (AFSs) combined with the radial basic function (RBF) is developed to evaluate liquefaction potentials of light loam induced by Tangshan earthquake in Tianjin area. The proposed system has strong self-adaptation and can dynamically adjust the number of hidden units through sample training under supervision. Six parameters related to earthquake and site conditions are selected as inputs and four outputs are designed to evaluate the extent of liquefaction according to code for seismic design of buildings. With the help of measurement data, it is shown that the AFSs-RBF network approach is able to predict liquefaction potentials and has a high success in training and testing for evaluating liquefaction classification of light loam.