Foundation stiffness identification for offshore platforms

Abstract

A system identification method for optimally estimating major uncertain parameters affecting the behavior of offshore structures, such as the foundation stiffness, using ambient or forced response measurements is presented. The main features of the system identification algorithm are: 1. 1. The generation of a quasi-structural model, expressed as a polynomial function analytically relating measured response parameters to the uncertain variables, in order to avoid expensive computer analysis using a detailed finite element model for the structure-foundation system. 2. 2. Optimization of the uncertain variables using a least-squares error minimization criterion. Two methods for obtaining numerical solutions, a generalized Newton-Raphson root finding algorithm and the Davidson-Fletcher-Powell gradient search method, are compared for accuracy and efficiency. An extensive sensitivity analysis is performed to: (a) assess and improve upon an algorithm originally applied to bridge structures, (b) study the influence of uncertainty in initially supplied parameter values on the accuracy of the iterative identification algorithm, and (c) investigate the effect of using partial modal information on the accuracy of parameter estimates. Finally, the system identification algorithm is applied to field data from an operating Gulf of Mexico caisson-type gas production platform to identify optimal values for the stiffnesses of a simple spring model for the foundation and to assess the stiffness contribution of grout filling the annulus between the caisson and an inner drive pile down to the mudline.