Uncertainty Quantification of fatigue design loads when compared with in-service measurements on ship-shaped offshore structures

Floating offshore structures are continuously subjected to wave loads and loading originating from operational activities. These loads lead to cracks growing in the structure, a process known as fatigue accumulation. The analysis of fatigue accumulation is subject to large uncertainties. This is related to the high sensitivity of fatigue accumulation with respect to the local stresses in the structure. Structural designs are analysed in advance to ensure they can fulfil the intended service life. To account for the uncertainties in the analysis, safety factors are applied. These safety factors on fatigue life can be as large as 10 for certain structural details. The spectral fatigue assessment procedure is most commonly used for the design of new structures. At several stages, models or assumptions are used which introduce uncertainty in the analysis process. In this research, the uncertainties in stresses which lead to fatigue accumulation have been examined. A number of parameters have been introduced to allow quantification of these effects on the fatigue accumulation process. The factors defined in this study relate to the contribution of the long-term wave environment, the sea state definition, the hydrodynamic and structural numerical models, weak and strong non-linear responses, the fatigue assessment method and loading-induced effects. The analyses are conducted using measurements from production units during operation in an offshore environment. Measurements from seven different units with a similar instrumentation setup have been used. These units are equipped with a variety of sensors including wave buoys or radar, motion sensors, loading computer and strain gauges. To ensure durability, strain sensors are installed on several non-critical elements at some distance from stress concentrations. At such locations, a uniform stress field governs which results in more accurate measurements. The sensors are positioned such that the local stresses are governed, as much as possible, by a single load effect. The variety of data sources available allows for the subdivision of the total uncertainty into multiple components as described above. This aids in understanding the importance of the separate assumptions made in the design process. Observations for individual units were conducted. By comparing the results obtained on multiple units, some generalized observations about the uncertainties were made. Uncertainties related to the long-term wave environment and short-term wave definition, were examined first. Dedicated wave buoys were used as sources of wave data. As alternative data source, information from hindcast models was used. These data sources were compared to obtain a general estimate of the uncertainty in long-term wave environments. The difference in fatigue accumulation when using these sources can be up to a factor of two. In cyclone-prone environments, the statistical uncertainty introduced by the presence of cyclone events is quite large and was assessed at 50%. Accurate sea state definitions are also quite important. Uncertainty arising from spectral definitions and wave spreading was quantified to be a factor of two on fatigue life. The importance of correctly modelling a confused sea state by using multiple wave systems varied strongly depending on the location. In the West-Africa region, the sea states are characterized by a strongly dominant swell. In other locations, a strongly confused sea state with multiple components was found and correct modelling of these components proved essential for an accurate fatigue assessment. The mismatch between the measured stress and those obtained using the numerical model is typically around 20% on stress for structural details subjected to global bending loads. However, on the structural details with wave pressure loads, deviations of the stress were found to be a factor of three. These deviations varied strongly between units and were mainly related to the linearisation of the intermittent wetting. The fatigue assessment method showed a higher accumulation rate than the directly assessed fatigue rate using strain gauges. A typical difference between them is 10%. However, the method becomes more conservative at higher fatigue accumulation rates. Therefore, the difference on the total accumulated fatigue life is larger, and typically a factor of two. Weak and strong non-linear effects have been quantified. Only minor weak nonlinear effects were found on the global loading effects. However, significant weak nonlinear effects in the side shell were found as a result of intermittent wetting. This is considered an important feature which requires more attention, especially for systems operating at constant draft. Strong non-linear effects, whipping and springing, were examined. Over the entire lifetime, the contribution of these effects did not exceed 15% at the analysed structural details. However, in unfavourable sea states, the contribution of whipping can be up to 20%. On large units, non-linear springing was found in sea states with low wave periods. In these sea states, the contribution of springing to the overall fatigue accumulation can exceed 100%. The main fatigue accumulation on cargo supporting structures, such as stringers, is related to loading cycles. The stress cycles at these details consist of contributions from cargo level variations in adjacent tanks, temperature-induced variations, wave-induced loads and secondary ballast and cargo loads. The variation of filling levels in adjacent tanks provides the largest contribution to the stress variation. However, the small additional stresses from the other sources result in a fatigue accumulation which is 50% larger than that based on cargo level variations alone. These effects are not fully accounted for in numerical analyses. Fatigue assessment procedures involve a large number of assumptions and modelling choices. In most cases, these assumptions introduce an additional safety margin on the fatigue life. However, a number of features in the applied loads were identified that can lead to higher stresses. Due to the high sensitivity of fatigue accumulation with respect to the stress level, the applied safety margins can be consumed quickly under disadvantageous conditions. Monitoring of an asset in service can help to quantify these uncertainties in the load prediction and can provide insight in the true safety margins of an offshore structure.