You are hereSecond-Order Load and Response Models for Floating Offshore Structures: Probabilistic Analysis and System Identification

Second-Order Load and Response Models for Floating Offshore Structures: Probabilistic Analysis and System Identification

Report No. : 
Todd C. Ude
December 1994
RMS16.pdf28.96 MB

Floating structures moored to the sea bottom provide a practical solution to the problem of producing oil and gas in deep water. Unlike more common fixed structures, the mooring systems of floaters can be designed such that the system's natural periods fall outside the
range of common wave periods. Due to nonlinearity in the wave to force transformation, however, oscillating forces can still arise in period ranges that will cause resonance in the structure. A second-order model of wave forces can be derived from a perturbation expansion of the linear form of the wave-struct ure interaction problem. These second-order, diffraction-analysis models have been used increasingly in recent years for modeling the linear and nonlinear wave forces on large volume structures.

The first objective of this work is the development of general methodologies that can be used for the efficient probabilistic analysis of these second-order models. The focus is on the prediction of response statistics of a deterministic system subject to a stationary,
random wave input (a seastate). For the study of this short-term uncertainty we combine the method of Kac and Siegert for finding statistical moments of a second-order system with the Hermite model for analysis of non-Gaussian extremes and fatigue damage. (Note that the Hermite model is a general, moment-based model that is in no way dependent upon the use of the Kac and Siegert method.) Combined with the power spectrum of output to reflect dynamic character, the moment-level description provides an accurate and flexible characterization of the random output.

Based on this analysis for fundamental structural responses such as mooring forces or horizontal displacements, we extend our analysis capabilities to include the prediction of fatigue damage accumulation and extreme values of response predicted by a second-order model. Integration of the conditional seastate analyses over the long-term variation of seastate parameters allows the development of a global view of the importance of nonlinear forces and the resulting non-Gaussian response. These analysis methods are demonstrated using a realistic model of an operating structure (the Snorre TLP). Results are compared to simulation, a robust but costly technique for generating similar results.

The second objective of this work involves the use of data (e.g. from model tests) in verifying and identifying models of the observed nonlinear behavior. First, assuming the diffraction model for wave forces to be correct, we use comparisons between predictions and model test observations to identify structural parameters such as damping. We find that the diffraction-based models can be calibrated to match observed responses with minimal effort, supporting the hypothesis that wave forces can be modeled via second-order diffraction .

Using a more traditional identification scheme, only the general framework of a second-order model is assumed and linear and quadratic transfer functions are fit to data via linear regression. The data requirements associated with robust estimation by this method are found to be significant. Use of the method on simulated data provides a benchmark for comparison of the performance resulting from observed-data identifications. Relatively greater success is obtained in fitting quadratic transfer functions for slow-drift responses, as compared to fitting transfer functions for sum-frequentcy tether-tension responses.