Abstract

This study examines the potential of multi-gene genetic programming (MGGP) based classification approach to evaluate liquefaction potential of soil in terms liquefaction index (LI) using a large database from post liquefaction cone penetration test (CPT) measurements and field manifestations. The database consists of CPT measurements; cone tip resistance (qc), friction ratio (Rf), vertical total stress (σv) and vertical effective stress of soil (σv′), seismic parameters; peak horizontal ground surface acceleration (amax) and earthquake moment magnitude (Mw), and the depth under consideration (z). The MGGP models (Model-I and Model-II) are developed for predicting occurrence and non-occurrence of liquefaction on basis of combination of above input parameters. The performance of the Model-I (95 %) is found to be more efficient compared to available artificial neural network model (91 %) and that of the Model-II (97 %) is found to be at par with the available support vector machine model (97 %) in terms of rate of successful prediction of liquefied and non-liquefied cases for testing data. Using an independent database of 96 cases, the overall classification accuracies are found to be 87 and 86 % for Model-I and Model-II respectively. Sensitivity analyses are made to identify the important parameters contributing to the prediction of LI.

Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call