Support Vector Classifiers for Prediction of Pile Foundation Performance in Liquefied Ground During Earthquakes

Abstract

Collapse of pile-supported structures is still observed in liquefiable soils after most major earthquakes and remains a continuing concern to the geotechnical engineering community. Current methods for pile design in liquefiable soils concentrate on a bending mechanism arising from lateral loads due to inertia and/or soil movement (kinematic loads). Recent investigations demonstrated that a pile or pile group can become laterally unstable (buckling instability/ bifurcation) under the axial load (due to the dead load) alone if the soil surrounding the pile liquefies in an earthquake. This is due to the liquefaction-induced elimination of the soil bracings and the governing mechanism is similar to Euler’s buckling of unsupported struts. Analysed are 26 cases of pile foundation performance in liquefiable soils giving emphasis to the buckling instability using Support Vector Machine (SVM) method. SVM has recently emerged as an elegant pattern recognition tool. This tool has been used to classify pile performance against buckling failure. Each of the case studies reported is represented by four parameters: Effective buckling length of pile (Leff), the allowable load on the pile (P), Euler’s elastic critical load of the pile (Pcr) and minimum radius of gyration of the pile (rmin). The performance of the developed SVM is 100%.