Reliability analysis of geostructures

Abstract

Geotechnical engineers always use experimental techniques for determination of newlinevarious parameters. Experimental techniques always give uncertainty. Reliability newlineanalysis is generally used to solve the problem of uncertainty. Hence, it is an newlineimportant task for geotechnical engineering practice. First order second moment newlinemethod (FOSM) is widely adopted by geotechnical engineers for determination of newlinereliability index. The main limitation of FOSM that it is time consuming method for newlineimplicit performance function. Researchers use response surface method (RSM), newlineArtificial Neural Network (ANN), multi-plane surface method, multi-plane tangent newlinesurface, etc. There are several drawbacks in the available models. Hence, advanced newlinemodels are required for overcoming the limitations of FOSM. This study employs newlineMinimax Probability Machine Regression (MPMR), Adaptive Neuro Fuzzy Inference System (ANFIS), Generalized Regression Neural Network(GRNN), Relevance Vector Machine(RVM), Gaussian Process Regression(GPR), Genetic newlinePrograming(GP), Multivariate Adaptive Regression Spline(MARS) to overcome the newlinelimitations of the FOSM model for determination of reliability index of shallow newlinefoundation, slope and retaining wall. ANFIS, GP, MPMR, GRNN, MARS, and GPR have been adopted as regression techniques. ANFIS uses neural network and fuzzy for prediction of output for a given input. It uses Membership Function (MF) to map the inputs. MARS adopts basis function for prediction of output. GRNN is developed based on kernel regression. Dichotomy classifier is used for constructing MPMR. Examples have been given to show the working procedure of ANFIS based FOSM, newlineGRNN based FOSM, RVM based FOSM, GPR based FOSM, GP based FOSM and MPMR based FOSM models. For development of models, the dataset has been divided into the two groups (training dataset and testing dataset). Training dataset has been used to construct the models. The developed models have been verified by testing dataset.