Reliability Analysis of Pile Foundation Using GMDH, GP and MARS

Risk and reliability are important aspects of geotechnical structures. Soil is highly variable. Pile tests are extremely expensive and labor-intensive. The paper proposes a reliable method to predict the bearing capacity of the pile using AI-based models like Multivariate Adaptive Regression Splines (MARS), Group method of data handling (GMDH) and Genetic programming (GP) using data of pile dynamic tests conducted on various sites in Indonesia. The performance of the model is ascertained using various performance parameters and are compared among themselves using Rank analysis and Taylor diagrams. The reliability indices of each model are calculated and compared. GP and MARS are concluded as robust models for estimating bearing capacity while performance of GMDH is not exciting.