A reliability study of an axially loaded jacket pile in sand for prediction of the failure probability in compression is presented. The study is based on full probabilistic modelling of axial load and axial capacity by their respective probability distributions. A first-order reliability method (FORM) is used for probability calculations.
Results by three novel axial capacity models are given, viz. the so-called UWA, ICP and NGI methods specified in ISO19902. Results by a fourth artificial method, whose capacity prediction is taken as the average of the predictions by the three other methods, are also given. Method-specific model uncertainty factors are applied, calibrated by means of a data base for piles in sand reported by UWA.
Emphasis in this paper is on presentation of results of the reliability analyses and on various sensitivity studies including two different failure criteria. The results illustrate that the novel design methods predict pile capacity with significant variability. The data base used for validation of the new design methods may therefore be judged inadequate, and the need for more large scale test results seems imperative to ensure a consistent safety level when these methods are used in design.
