Response Based Analysis (RBA) is an advanced method for the prediction of long term distributions of critical responses in offshore floating systems. For complex non-linear systems such as flexible risers, RBA requires time domain simulations that form the core data to which probabilistic models are applied. Because RBA requires significantly larger amounts of data than traditional short term analysis approaches, running the required number of simulations in the time domain can quickly become unfeasible if the system’s physics being modelled are exceedingly complex. In addition, flexible risers are complex composite structures with highly dynamic, non-linear responses which further limit the feasibility of application of the RBA process to these systems.
As an alternative, frequency domain solvers, such as that used in the OrcaFlex software, are potential substitutes for portions of datasets due to their processing times being significantly faster than time domain solvers.
A comparison of extreme responses generated by frequency and time domain solvers was performed over the duration of two storms. An upper threshold limit for the frequency domain’s accuracy was found by comparing the differences of the two solver’s responses as the storm progressed; where the differences became too large the threshold limit was set. For environmental conditions smaller than this threshold, the frequency domain solver may provide a quicker method for predicting the riser responses. Conditions that exceed this threshold require full time domain analysis for accurate responses to be generated.
Limitations of the frequency domain solvers include their reduced ability to deal with non-linear mechanics such as bending/curvature responses. As a result, curvature component results from the frequency domain are limited in their direct usability, especially when exposed to more extreme metocean conditions and locations along the riser that are subject to larger curvature (generally where risers are connected to structures with greater stiffness).
Although these limitations exist, the frequency domain solver may still provide reasonable insight into metocean conditions that potentially cause extreme responses.
A method is proposed for the use of both frequency and time domain simulations in the flexible riser flowline RBA process. Screening, filtering and ‘stitching’ methods utilizing the speed of the frequency domain solver are presented in order to compensate for the time domain’s extensive computation times. The proposed method of stitching, when applied to an example storm history, required 39% of the processing time when using only the time domain solver.