Large-scale model tests of a tensioned steel riser were performed at Hanoytangen outside Bergen, Norway in 1997. The length of the model was 90 m and the diameter was 3 cm. The information from these tests consists of measured bending strains, tension, flow speed and all relevant riser data. In this work, this information is reexamined in an attempt to improve our understanding of vortex-induced vibrations (VIV) for cases with very high order of responding modes. The aim is in particular to study the relative importance of in-line (IL) and cross-flow (CF) vibrations for fatigue damage accumulation. It is shown that fatigue damage is proportional to U2m+1 (U is the flow velocity) when the modes are dominated by tension. When bending controls the modes, the fatigue damage is proportional to Um+1. A linear SN-curve with slope parameter m = 3 is used. The Hanoytangen riser fatigue damage goes as U-7 for the lowest velocities and U-4 for the highest current velocities. Based on the Hanoytangen data, it seems that the transition velocity between the tension and the bending-stiffhess-dominated regions is at the current velocity that gives response at a mode number where a tensioned string and an untensioned beam have equal eigenfrequencies. IL response has a significant contribution to fatigue for cases dominated by the lowest modes. The reason is that IL oscillations will take place at double the frequency of those in CF. For a tension-controlled case, this corresponds to a mode with half the wavelength, while a bending-controlled case will tend to have a wavelength ratio of root 2. Since the curvature for a given amplitude increases with the inverse modal wavelength squared, fatigue from IL tends to dominate for cases with tension-controlled modes (low current speed), while CF will dominate for bending-controlled modes (high current speed). This tendency is clearly seen in the experimental data for both CF and IL responses. Fatigue damage is calculated directly from the measured data and compared to results found by using a computer program based on a semi-empirical method. The program is able to calculate response frequencies, response amplitudes and fatigue damage. The method is limited to account for CF vibrations only. Computed results are compared to measurements. The agreement is in general good, but some discrepancy is seen for cases with high current velocity. In these cases, the real response contains high mode orders (above 25) and tends to have a stochastic nature, while the VIV-analysis model assumes that the response takes place at a limited number of discrete frequencies. Further development of the empirical model is therefore needed. (c) 2005 Elsevier Ltd. All rights reserved.
