Turret structures within FPSO's are subjected to a variety of repeated loads, due to the harsh environment they operate in during their service live, meaning that fatigue is a governing limit state in turret structural design. Turret structural components are connected by arc-welded joints which are considered particularly susceptible to fatigue damage. Due to the complexity of the structure and its loading, welded details may be subjected to a multiaxial stress state. Currently fatigue design of turret structures is predominantly based on a uniaxial fatigue criteria assuming governing mode I (i.e. normal stresses). This design approach can be non-conservative for welded details subjected to a multiaxial stress state, especially when these are non-proportional (i.e. out of phase). The estimation of multiaxial fatigue live for details subjected to a multiaxial stress state is still an extremely complex task. There is still a discrepancy in obtained multiaxial fatigue live between different design rules (i.e. as presented in ISSC), meaning that future work on this topic is required.
Due to the size and complexity of turret structures, identification of welded details (i.e. in the order of hundreds) subjected to a multiaxial (non)-proportional stress state is a rather complicated and laborious task. This thesis proposes a new screening method that identifies sensitive locations where multiaxiality occurs, either geometry or loading induced. Component stresses (i.e. Mode I and III) are determined from finite element models using a mesh-insensitive structural stress method. The stress state (i.e. multiaxiality and proportionality) of these stress components is determined using the parameters of an ellipse that encloses the stress data (i.e. component stresses) in 2-dimensional stress space. Making this a practical and efficient method to identify sensitive locations within the turret where multiaxiality occurs.
For the multiaxial fatigue damage calculations of welded details subjected to a non-proportional multiaxial stress state, the accumulative Moment of Load Path (MLP) concept is used. Within this concept the multiaxial fatigue damage for any given non-proportional load path is assumed to consist of two parts. The first part can be considered damage due to the effective stress range Δσ_e (i.e. stress due to direct path), and the second is the "load path non-proportionality" fatigue damage due to excursion of the reference load path. By implementing the MLP-based method as part of the path dependent maximum range (PRMD) cycle counting procedure, half cycles and their corresponding MLP-based equivalent stresses ranges are computed. Given the MLP-based stress distributions from PDMR cycle counting, the well know Palmgren-Minor rule is used to determine the accumulation of fatigue damage considering a proper fatigue resistance curve. By implementing both the screening and the proposed multiaxial fatigue damage method onto a relatively simple tube-to-flange connection, a comparison study is used to determine whether the screening method is capable of identifying sensitive locations that may be susceptible to multiaxial fatigue. For five considered load scenarios, the screening method showed to give relatively similar results with respect to the actual fatigue damage calculation, making it a suitable structural screening method. Using the same tube-to-flange connection the multiaxial fatigue damage is calculated based on DNVGL and compared with those calculated using the MLP-based concept. For uniaxial proportional loading scenarios similar fatigue damages are observed, however for the non-proportional load scenarios the MLP-based method gives significant higher fatigue damage results w.r.t DNVGL. Performing structural screening on complex structural systems like the turret considering each load signal during its service live would be very computationally expensive. Therefore five load scenarios are considered to be sufficient to perform the structural screening. For this thesis a certain domain of the turret of the Aoka Mizu vessel is screened using the defined screening load scenarios. Within the scope (i.e. evaluated domain) details resulting in relative high fatigue damage due to high levels of non-proportionality are not encountered. Even though details exist with high non-proportionality factors the stress ranges of these details are usually relatively small compared to details subjected to a dominant uni-axial stress state.