Predicting Liquefaction-Induced Lateral Spreading by Using Neural Network and Neuro-Fuzzy Techniques

The prediction of lateral spreading is an important task because of the complexities of lateral-spreading behavior. The aim of this work is to improve an accurate liquefaction-induced lateral-spreading prediction by using multiple regression methods, such as multilinear regression (MLR), multilayer perceptrons (MLPs), and the adaptive neuro-fuzzy inference system (ANFIS). Predictions of lateral spreading from the developed MLR, MLP, and ANFIS models in tractable (susceptible) equation form are obtained and compared with the value predicted using traditional methods. Principal-component analysis is used to evaluate the effects of each input variable on the lateral spreading. On the basis of the comparisons, it is found that the MLP is better than the ANFIS, MLR, and Youd equation for estimating maximum lateral displacement of free-face conditions. For gently sloping ground conditions, however, similar results are obtained with MLP and ANFIS, which are better than the MLR and Youd equation. The MLP model was also tested with data obtained from Adapazari, Turkey, to estimate total lateral displacement.